Development and validation of a comprehensive microglia profiler to obtain an in-depth 3D morphometric analysis of various cellular parameters
To investigate potential HF-induced changes in microglial and astrocyte morphology, as well as pro-inflammatory cytokines levels and astrocyte markers, we used our established model of HF (Fig. 1a) and performed morphometric and genetic assessments at different time points post-surgery (Fig. 1b). In general, assessment of microglia staining during neuroinflammatory processes has relied on the evaluation of IBA1 immunoreactivity intensity and density within the tissue. However, this approach is insufficient to provide critical information regarding various morphometric microglial parameters such as degree of ramification, somatic volume and overall cellular complexity. Thus, to obtain a detailed quantitative morphometric analyses of glia cells in HF, we used our novel glial profiler based on the IMARIS software to perform 3D reconstruction of microglia and astrocytes. First, we probed for a potential HF-induced microglial proliferation. We did not find significant differences in total microglial cell numbers between sham and HF rats in both the PVN (sham 14w: 34.3 ± 4 cells vs HF 14w: 35.2 ± 4 cells, n.s., n=4) and the CeA (sham 14w: 34.7 ± 4 cells vs. HF 14w: 35.1 ± 3 cells, n.s., n=4) (Fig. 1c, d), indicative that during HF there is no microglial proliferation within these brain regions at 14 weeks post-surgery.
We then used a more detailed approach to assess changes in individually identified microglia cell surface area, cell volume, filament length, branches and morphometric complexity (Fig. 2a). In brief, we took 50µm z-stack confocal images (16-bit, 1 µm steps, 40x objective) of brain slices stained with IBA1 and exported czi files for further analysis in IMARIS. Incompletely stained microglia were not included into our analysis (detailed description of IMARIS-based analysis can be found in ‘Materials and Methods’). This method allows an unbiased high throughput analysis of various cellular features and is a universally applicable tool for morphometric analysis of different cell types under various conditions. To assess whether the sham surgery itself resulted in microglia morphometric alterations over time, we perfused naïve (20 weeks old, n=2) and sham rats (8-, 14- and 16-weeks post-surgery, n=4/group). We stained brain sections containing the PVN and the CeA with IBA1 and analyzed microglial profiles using IMARIS. We did not observe any significant changes in the PVN or CeA of sham rats in any of these parameters either when compared to naive rats (20w), or as a function of time after the surgical procedure (8w, 14w or 16w, Fig. 3a-d). Therefore, we decided to pool together the three sham groups for further analysis.
HF-induced a time-dependent morphometric microglial changes in the PVN that are already evident at 8 weeks post-surgery
To study HF-induced microglial changes in the PVN and CeA over time, we perfused sham and HF rats at 8, 14 and 16 weeks after the surgery (n=4/group) and assessed microglial surface area, cell volume, filament length, microglial branches, microglial segments, filament terminals and IBA1 density. Representative confocal images of PVN microglia of sham and HF rats 16-weeks post-surgery are shown in Fig. 4a, the anatomical location of the CeA is shown in Fig. 4b. As depicted in Fig. 4b, we found significant morphometric microglial changes already at 8 weeks post-surgery, which progressed as a function of time. We found a progressive decrease in microglial surface area (0.9%, n.s., 12.3%, p=0.0051 and 17.1%, p<0.0001 at 8, 14, and 16 weeks, respectively, one-way ANOVA, F=11.34, p<0.0001); a progressive decrease in microglial cell volume (7.0%, p=0.0475, 27.5%, p<0.0001 and 34.0%, p<0.0001 at 8, 14, and 16 weeks, respectively, one-way ANOVA, F=48.21, p<0.0001); a decrease in filament length (10.6%, p=0.0002, 12.5%, p<0.0001 and 26.1%, p<0.0001 at 8, 14, and 16 weeks, respectively, one-way ANOVA, F=22.46, p<0.0001); a progressive reduction in the number of microglial branches (17.8%, p<0.0001, 21.0%, p<0.0001 and 26.6%, p<0.0001 at 8, 14, and 16 weeks, respectively, one-way ANOVA, F=19.33, p<0.0001); a progressive reduction in the number of microglial segments (24.2%, p<0.0001, 26.8%, p<0.0001 and 32.2%, p<0.0001 at 8, 14, and 16 weeks, respectively, one-way ANOVA, F=40.20, p<0.0001); a reduction in the number of filament terminals (22.3%, p<0.0001, 26.0%, p<0.0001 and 28.1%, p<0.0001 at 8, 14, and 16 weeks, respectively, one-way ANOVA, F=25.9, p<0.0001) and a progressive increase in IBA1 intensity (17.4%, p<0.0001, 34.4%, p<0.0001 and 34.6%, p<0.0001 at 8, 14, and 16 weeks, respectively, one-way ANOVA, F=89.41, p<0.0001. These findings indicate that HF-induced microglial morphological changes in the PVN are evident as early as 8 weeks after the myocardial infarction.
Next, to rule out that HF-induced changes in microglia morphology was a diffuse phenomenon affecting the whole brain, we chose the primary somatosensory cortex (barrel field, S1BF, Fig. 5a-c) as a functionally-unrelated control region and analyzed IBA1-labeled microglia in this structure. We did not observe any significant microglial changes within this region, suggesting that, at least at this stage cortical brain areas do not display significant changes in microglial morphology.
Lastly, to determine whether the observed microglial changes were dependent on the severity of the disease, we included an additional control group of rats that underwent the myocardial infarction surgery (and used at 16 weeks post-surgery), but that developed only a mild form of functional heart failure, indicated by an EF>50% (mean EF: 64.9 ± 2.1%, n=5). We found that a mild functional HF was not associated to any significant changes of microglial morphology, given that all our tested parameters were similar between our two groups (sham 16w vs. HF EF>50% 16w, Fig. 6a-c). Moreover, this additional group allowed us to perform a correlative analysis of the degree of microglia changes as a function of the severity of HF (i.e., as a function of EF value). For this analysis we selected three representative and key microglia morphometric endpoints (cell volume, filament length and number of microglial branches) from 3 experimental groups: sham rats (n=4, 16w) HF <50% EF (n=4, 16w) and HF>50 %EF (n=5, 16w). As summarized in Figure 6d, we found a strong and significant correlation for all three morphometric parameters and EF, supporting the notion that morphometric changes in microglia during HF are dependent on the severity of the disease.
HF-induced delayed morphometric microglial changes in the CeA that correlated with the severity of the disease
Representative confocal images of CeA microglia of sham and HF rats 16-weeks post-surgery are shown in Fig. 7a. In stark contrast to the PVN, except for a slight decrease in filament length, we did not detect any significant changes in microglia morphology 8 weeks after surgery, indicating that HF-induced changes in microglial morphology do not affect the CeA at this point in time. However, some changes in microglial morphometric parameters emerged at 14 weeks, which continued to progress at week 16. We found a decrease in microglial surface area (2.4%, n.s. and 15.6%, p=0.0063 at 14 and 16 weeks, respectively, one-way ANOVA, F=4.821, p=0.034); an initial increase by 21.4%, p<0.0001 followed by a later decrease by 12.7%, p=0.028 in microglial cell volume at 14 and 16 weeks, respectively, one-way ANOVA, F=12.09, p<0.0001; a progressive decrease in filament length (12.0%, p=0.014, 10.2%, p=0.041 and 28.3%, p<0.0001 at 8, 14 and 16 weeks, respectively, one-way ANOVA, F=10.60, p<0.0001); a progressive reduction in the number of microglial branches (10.0%, p=0.0244 and 15.1%, p=0.012 at 14 and 16 weeks, respectively, one-way ANOVA, F=5.62, p=0.0001); a reduction in the number of microglial segments (12.1%, p=0.01 and 18.5%, p=0.027 at 14 and 16 weeks, respectively, one-way ANOVA, F=6.442, p<0.0001); a progressive reduction in the number of filament terminals (15.3%, p=0.01 and 21.0%, p<0.0001 at 14 and 16 weeks, respectively, one-way ANOVA, F=9.87, p<0.0001) and a progressive increase in IBA1 intensity (11.1%, p<0.0001 and 25.5%, p<0.001 at 14 and 16 weeks, respectively, one-way ANOVA, F=51.28, p<0.0001, Fig. 7b, c). It is important to note that while we observed an initial increase of microglial cell volume at 14 weeks-post surgery in CeA, at 16 weeks post-surgery the cell volume was significantly decreased compared to the sham group. These findings suggest that during HF, microglia undergo cellular/somatic swelling prior to the retraction of processes and decrease in cellular volume.
Similarly to what we observed in the PVN, microglia morphometric parameters were not changes in rats displaying a mild form of HF (i.e., EF>50%; n=5, see results summarized in Table 1). Importantly, we also found key CeA microglia morphometric endpoints (cell volume, filament length and microglial branches) to strongly correlate with the severity of the disease (Fig. 7d). Taken together, these results indicate that CeA microglia undergo morphometric changes during HF, which occurred with a delayed time course compared to PVN microglia and which also correlated to the severity of the disease. The complete overview of all individual values from all groups can be found in Table 1.
HF results in microglial deramification and somatic swelling in both PVN and CeA
As stated above, pro-inflammatory microglia have been demonstrated to undergo deramification, a process where microglia retract their processes, lose microglial complexity and release inflammatory cytokines (42, 52). Thus, to further investigate microglial cell morphometric changes during HF in the PVN and CeA, we performed a Sholl analysis of individually 3D-reconstructed microglial cells in each experimental group. To this end, we superimposed spheres of increasing radius (1 µm increase in radius per step, Fig. 8a) starting at the center of the soma, and measured the number of process intersections that each sphere encountered. We found that in the PVN (Fig. 8b) microglia displayed a significant loss of complexity (indicated by a significantly reduced average number of total Sholl intersections) that occurred in a time-dependent manner with changes being already detectable 8 weeks post-surgery, and becoming progressively more deramified at 14- and 16-weeks post-surgery (Fig. 8b, 11.1%, p=0.008, 17.1%, p<0.0001 and 30.8%, p<0.0001 at 8, 14, and 16 weeks, respectively, one-way ANOVA, F = 36.3, p<0.0001). In the CeA (Fig. 8c), we did not observe such changes at 8- or 14-weeks post-surgery. However, 16-weeks post-surgery, microglia in the CeA displayed significant deramification (sham: 641.3 ± 34 vs. HF 16w: 534.7 ± 25, p=0.036, one-way ANOVA, F=7.42, p<0.0001). A recent two-dimensional high throughput study on microglia suggested that detailed morphometric analyses of microglia might be more informative and yielded valuable information leading the authors to coin the terms ‘low and high-activity’ microglia (53). However, given that microglia are highly active cells both at ramified and deramified cellular states, we chose a more descriptive/neutral classification into ramified and deramified (pro-inflammatory) microglia. To quantify the degree of overall microglia deramification, we first determined the peak number of Sholl intersections (i.e., the highest numeric value of a sphere intersecting with a microglial process) per individual microglial cell, which in our entire sampled microglia cell population ranged from 0 to 64 (the higher the number, the more ramified the microglial structure). Using a semi-arbitrary and conservative threshold of <10 to categorize a microglial cell as pro-inflammatory, we found 32.5% deramified microglia in the PVN (Fig. 8d) and 30.9% in the CeA (Fig.8e) of sham rats 14 weeks post-surgery In the HF rats, we found a significant increase in the number of deramified microglia 8 weeks post-surgery in the PVN (HF 14w: 41.4% vs. sham: 32.5%, p=0.0125, one-way ANOVA, F=25.70, p<0.0001), but not the CeA (31.8%, n.s.). However, 14 weeks post-surgery we found a significant increase in the number of deramified microglia in the CeA (38.2%, p=0.0201, one-way ANOVA, F=12.12, p<0.0001), while the number of deramified microglia in the PVN had further increased (55.9%, p=0.001). In the CeA, the percentage of deramified microglia had further increased 16 weeks post-surgery (HF 16w: 44.0%, p=0.0163), which was not the case for the PVN (60.25%, n.s.).
Finally, we analyzed the Sholl distribution curves for sham and HF in the PVN and the CeA at 16-weeks post-surgery and also compared microglia between PVN and CeA in sham animals (Fig. 8f-h). We found significant differences for both the PVN (two-way ANOVA, group: F (1, 62)=96.48, p<0.0001, Fig. 8f) and the CeA (two-way ANOVA, group: F (1, 62)=83.77, p<0.0001, Fig. 8g). In the PVN the mean peak number of Sholl intersections was reached at 21 µm distance from the soma for both HF and sham animals, and the mean number of intersections at the peak was significantly different between those groups (sham: 18.5 ± 1.2 intersections vs. HF 16w: 10.2 ± 0.9 intersections, p<0.0001). In the CeA, this peak was reached at 22 µm for the sham group and 21 µm for the HF group and was also significantly different between the two groups (sham: 21.9 ± 1.6 intersections vs. HF 16w: 13.9 ± 1.3 intersections, p<0.0001). Interestingly, under control conditions microglial cells in the CeA were significantly more complex than in the PVN (two-way ANOVA, group: F (1, 62)=74.13, p<0.0001, Fig. 8h), highlighting the morphological heterogeneity among microglia within different brain regions. These findings suggest that microglia in both PVN and CeA undergo morphometric changes and deramification during HF.
A recent study highlighted that in addition to deramification, pro-inflammatory microglia display somatic swelling (53), a process thought to coincide with the release of pro-inflammatory cytokines, especially in neurodegeneration (54). Thus, to investigate whether HF surgery resulted in somatic microglia swelling, we used our 3D profiler to calculate the somatic volume of individual microglial cells. We found that the average microglial soma volume of sham rats 14 weeks post-surgery was 554.3 ± 21 µm3 for the PVN and 529.0 ± 41 µm3 for the CeA (Fig. 9a-c). We found a time-dependent increase in somatic volume of PVN microglia: 28.2%, p=0.0025 and 43.1%, p<0.0001 at 14 and 16 weeks, respectively, compared to the respective sham group, one-way ANOVA, F=12.48, p<0.0001. In the CeA, we found even more time-dependent increases in somatic volume: 41.3%, p<0.0001 and 51.2%, p<0.0001 at 14 and 16 weeks, respectively, compared to the respective sham group, one-way ANOVA, F=20.37, p<0.0001.
We next sought to investigate whether somatic swelling and microglial deramification were correlated processes or whether they occurred independently in separate microglial subpopulations. Therefore, we correlated soma volume and the total number of Sholl intersections for each microglial cell (Fig. 9d-g). We found that both in sham and HF rats, somatic swelling and deramification were highly correlated processes and this held true for both PVN and CeA (p<0.0001 for all cases). Interestingly, we observed apparent decreases in the slope of the best-fit non-linear regression line (Fig. 9d-g, red lines) in HF compared to sham rats, in both brain regions, which might be indicative of the less polarized (i.e., more homogenous) microglial cell population spectrum in HF rats, resulting from the shift towards a pro-inflammatory microglial stage. It is important to note that although we found CeA microglia to have a smaller average somata than PVN microglia in sham rats, they surpassed PVN microglia in soma volume at 14 weeks post-surgery. This finding suggests that although somatic swelling and deramification are correlated phenomena, somatic swelling might be initiated earlier and could potentially explain the initial increase in total cell volume in CeA microglia at this time (Fig. 7b, cell volume, HF 14w).
HF induces a morphological A1 astrocyte phenotype in both PVN and CeA
During injury and disease, microglia and astrocytes display intricate interactions that may lead either to neuronal survival or neuronal loss. A recent study showed that activated (pro-inflammatory) microglia induce a neurotoxic A1 subtype of astrocytes, which secrete a currently unknown neurotoxin that results in neuronal cell death (48). A1 astrocytes can be discriminated from the neuroprotective A2 astrocytes not only by the upregulation of genetic markers (47-49, 55), but also by distinct morphological changes, which appear to be similar to those of microglia during neuroinflammation (56). While retraction of astrocytic processes, hypertrophy and gliosis has been described for astrocytes under various conditions (57), no study – to the best of our knowledge – comprehensively addressed alterations in astrocyte morphology during chronic neuroinflammation using a three-dimensional reconstruction profiler. Given that neuroinflammation-induced changes in astrocytes seem to generally follow those observed in microglia (47, 48), we chose to investigate astrocytic changes at 14 weeks post-surgery.
To investigate whether HF results in the induction of the A1 astrocytic phenotype, we performed 3D IMARIS analysis of astrocytes (Fig. 10a) that were immunohistochemically identified by their expression of both GFAP and glutamine synthetase (GluSyn) (Fig. 10b, e). We specifically chose a combination of these two markers because they allowed us to perform a detailed analysis of astrocytic somata and processes, given that GFAP stains mostly processes, but not the soma, while GluSyn predominantly stains the astrocyte soma, but not processes (see Fig. 10j). In HF rats, we found significant morphological changes in both PVN and CeA astrocytes 14 weeks post-surgery (n=4 rats/group) that included a decrease in surface area (17.2%, p=0.0062 and 19.1%, p=0.022, for PVN and CeA, respectively), cell volume (24.4%, p=0.001 and 23.2%, p=0.001, for PVN and CeA, respectively), filament length (19.4%, p=0.0052 and 24.6%, p=0.0033, for PVN and CeA, respectively) and an increase in soma volume (30.8%, p<0.0001 and 18.8%, p=0.026, for PVN and CeA, respectively, Fig. 10c, f). Furthermore, we found that astrocytes of HF rats displayed a significant loss of process complexity, as shown by a significant decrease in the total number of Sholl intersections per astrocyte both in the PVN (18.0%) and the CeA (27.8%, p<0.0001 in both cases, Fig. 10d, g). In addition, the Sholl distribution analysis revealed significant changes in astrocyte complexity 14-weeks post-surgery for both the PVN (two-way ANOVA, group: F (1, 81)=84.32, p<0.0001, Fig. 10h) and CeA (two-way ANOVA, group: F (1, 62)=70.92, p<0.0001, Fig. 10i). Fig. 10j shows an isolated PVN astrocyte of a HF rat and subsequent surface reconstruction via IMARIS.
HF induces expression of genes associated with neuroinflammation and A1 astrocyte phenotype in both PVN and CeA
To determine whether the microglial/astrocyte morphometric changes observed in HF rats also corresponded with a genetic profile associated to neuroinflammation and/or a shift to an A1 astrocyte phenotype, we performed qPCR of mRNA transcripts of various genes classically associated to neuroinflammation, as well as several A2 and A1 astrocyte-related genes. We analyzed mRNA levels of IBA1 and GFAP, cytokines (TNF-a, IL-1b and IL-6), A1 astrocyte markers (Serping1 and C3) and A2 astrocyte markers (Tm4sf1 and Sphk1) in micropunches obtained from the PVN and CeA of sham and HF rats at the same time points (8w and 14w) at which the morphometric studies were performed.
In the PVN, we found significant changes in tested genes already at 8 weeks post-surgery (n=5 per group, pooled): an increase in IBA1 (1.344-fold, p=0.0006), cytokines (IL-1b: 3.51-fold, p=0.0163; IL-6: 1.522-fold, p=0.0067 and TNF-a: 3.117-fold, p=0.0016) and A1 astrocyte markers (Serping1: 2.767-fold, p=0.017 and C3: 1.127-fold, p=0.0323) as well as a significant decrease in GFAP (-1.583-fold, p=0.0006) and A2 astrocyte markers (Tm4sf1: -1.373-fold, p=0.0005 and Sphk1: -1.17-fold, p=0.0005, Fig. 11a). When we analyzed the genes again 14 weeks post-surgery (Fig. 11b) (n=4/group, pooled), the qPCR of PVN tissue yielded comparable results: IBA1: 1.683-fold, p=0.0089; GFAP: -1.563, p=0.0006; IL-1b: 3.533-fold, p=0.0134; IL-6: 1.62-fold, p=0.0045; TNF-a: 2.433-fold, p=0.0344; Serping1: 4.2-fold, p=0.0042; C3: 1.387-fold, p=0.0237; Tm4sf1: -1.897, p<0.0001, Sphk1: -1.333, p=0.0012).
In the CeA we found only GFAP (-1.249-fold, p=0.0004), IL-1b (2.383-fold, p=0.0291) and TNF-a (2.367-fold, p=0.004) to be significantly altered 8 weeks post-surgery (Fig. 11c). However, 14 weeks post-surgery (Fig. 11d) we found significant differences in all tested genes, which were comparable to those observed in the PVN: IBA1 (1.433-fold, p=0.0039), GFAP (-1.323-fold, p=0.0002), IL-1b (2.893-fold, p=0.0066), IL-6 (1.757-fold, p=0.0035), TNF-a (2.72-fold, p<0.0001), Serping1 (5.033-fold, p=0.0009), C3 (1.243-fold, p=0.0404), Tm4sf1 (-1.738-fold, p=0.0004) and Sphk1 (-1.57-fold, p=0.0009). We did not observe significant changes either at 8 weeks or 14 weeks post-surgery in the prelimbic cortex (PLC, Fig. 11e, f), suggesting that this cortical area might not be affected by HF at this point in time.