Surgical Protocol and Ischemic Duration Determine Myeloid Cell Features and Locomotor Deficits in Preclinical Transient Middle Cerebral Artery Occlusion

Abstract

Ischemic stroke remains a leading cause of disability, despite increasing patient eligibility for undergoing vessel recanalization. Ischemia/reperfusion injury (I/RI), the inflammatory cascade triggered by ischemia and subsequent reperfusion, contributes to infarct evolution and functional outcomes. However, investigation of mechanisms underlying I/RI is hindered by variable outcomes reported across preclinical studies, which can largely be attributed to differences in the stroke models used. Although the transient middle cerebral artery occlusion (tMCAO) model has been widely adopted to recapitulate I/RI, interchangeable use of the Longa and Koizumi surgical protocols and a range of durations of occlusion have resulted in inconsistent reports of pathological and functional outcomes. To evaluate the effects of tMCAO model variability, we employed the Longa and Koizumi tMCAO protocols across increasing ischemic durations. Using multimodal imaging and behavioral testing, we report that infarct pathology, locomotor deficits, and innate immune responses were significantly influenced by ischemic duration and surgical protocol at 24 and 72 hours post-tMCAO. Myeloid cell morphology and spatial relationships were the most robustly affected cellular features but were differentially affected by ischemic duration and surgical protocol. These findings provide a comprehensive assessment of the consequences of variable tMCAO dynamics to guide experimental design for future preclinical studies.

BACKGROUND

Ischemic stroke is a leading cause of morbidity and mortality worldwide [1]. Large vessel occlusions (LVOs), in which a clot occludes a major cerebral artery to induce ischemia, are the most morbid subtype of ischemic stroke [2]. LVOs are managed by recanalization of the blocked vessel within 24 hours of stroke onset to facilitate reperfusion. Although eligibility criteria for recanalization have been continuously expanding, >50% of LVO patients still experience significant long-term disability [3–6]. These sustained functional deficits – despite maximal treatment of ischemia – drive the continued exploration of mechanisms that define stroke severity and have the potential to serve as future therapeutic targets.

Ischemia/reperfusion injury (I/RI), the inflammatory response induced by ischemia and continued throughout reperfusion, significantly contributes to infarct evolution and functional outcomes [7]. The acute stages of I/RI are predominantly mediated by myeloid cells – polymorphonuclear neutrophils (PMNs), bone marrow-derived macrophages (MΦ), and microglia (MG) [8, 9]. Given that the post-stroke inflammatory response is highly conserved between mice and humans, mouse models are routinely used to mimic stroke pathology [10]. In particular, the transient middle cerebral artery occlusion (tMCAO) model has been widely adopted to simulate LVO I/RI, as it recapitulates ischemia and reperfusion by inserting and subsequently withdrawing an intraluminal filament. However, pathological, functional, and inflammatory measures of stroke severity in the tMCAO model have been observed to vary widely across preclinical studies. A previous meta-analysis found that average infarct volumes range from 9-100 mm³ and the pathological involvement of striatal and cortical territories differed dramatically between cohorts [11]. Further, the timing and extent of post-tMCAO myeloid cell recruitment is also unclear because of conflicting reports of immune cell dynamics.

This lack of consensus can be largely attributed to a high degree of variability in the tMCAO methodology in terms of ischemia and reperfusion dynamics. The original intraluminal tMCAO technique (Koizumi) as well as a subsequent modified protocol (Longa) are both used interchangeably in preclinical models, despite differing in the route of occlusion, and therefore, the degree of reperfusion achieved [12, 13]. The Koizumi protocol uses the common carotid artery (CCA), while the Longa protocol uses the external carotid artery (ECA) for filament insertion and withdrawal. Although both techniques result in acute focal ischemia, the loss of the CCA in the Koizumi protocol also induces chronic hyporeperfusion that further contributes to longitudinal infarct evolution and outcomes. This is in contrast to the complete reperfusion achieved using the Longa protocol. The duration of ischemia is also highly inconsistent across studies, ranging from 15 minutes to multiple hours. These variations in surgical approach have resulted in a lack of clear definition of the relevance of ischemia and reperfusion dynamics in mediating post-tMCAO inflammatory outcomes, complicating the selection of an appropriate model for preclinical studies.

Differences in infarct severity have previously been established between the Koizumi and Longa models of tMCAO, emphasizing the importance of understanding how complete and incomplete reperfusion affect pathology and outcomes. However, it remains unknown whether this effect of surgical protocol on infarct severity correlates with the extent of the immune response, and ultimately locomotor function. Further, the effect of ischemic duration on the outcomes observed in the Longa and Koizumi protocols has never been evaluated despite the broad range of ischemic windows employed. Therefore, we sought to compare how the Koizumi and Longa protocols at varying ischemic durations affect pathological, inflammatory, and functional outcomes. Using serial perfusion imaging and behavioral testing in myeloid reporter mice, this study reports that acute infarct topology, behavior, and myeloid cell features are more strongly modulated by ischemic duration than by reperfusion efficacy. Specifically, MG/MΦ and PMN morphology, and spatial relationships between PMNs and MG/MΦ were most profoundly affected. These data highlight the need to consider ischemic duration and degree of reperfusion in the design of future preclinical stroke studies.

METHODS

Mice

All procedures were approved by the Institutional Animal Care and Use Committee (Animal Welfare Assurance No A3011-0) at Stony Brook University. All experiments were conducted according to the Institutional Animal Care and Use Committee and guidelines from the National Institute of Health, and in accordance with the ARRIVE guidelines. Mice were housed in the institutional animal facility under a 12-hour light/dark cycle, and all husbandry was performed by the Stony Brook Division of Laboratory Animal Resources. Food and water were provided ad libitum. The Ly6G-TdT x Csf1r-EGFP strain was derived by crossing Catchup (C57BL/6 Catchup^IVM, Ly6g-cre Rosa26-CAG-tdTomato) mice with MacGreen (C57BL/6-CSF1R-EGFP) mice in house to yield heterozygotes for each fluorophore [14–16]. Catchup mice were generously provided by Drs David Sullivan and William Muller at Northwestern University. MacGreen mice were acquired from Jackson Laboratories (Strain #018549, B6.Cg-Tg(Csf1r-EGFP)1Hume/J). Both reporters have been used extensively to study inflammation [14, 17–20]. Genotypes were confirmed via polymerase chain reaction. Equal numbers of male and female Ly6G-TdT x Csf1r-EGFP heterozygotes aged 3 to 6 months were used for all tMCAO experiments. Age-matched Ly6G-TdT heterozygotes were used for sham experiments according to mouse availability. Mice weighing <20g or >35 g were excluded from analyses. In total, 295 mice were used.

Transient Middle Cerebral Artery Occlusion (tMCAO) Model

Surgical procedures were performed following the IMPROVE guidelines [21]. Mice were randomized to either the Longa (complete reperfusion) or Koizumi (incomplete reperfusion) tMCAO protocol, as previously described, and were further randomized to 30, 60, or 90 minutes of ischemia [12, 13]. Mice received preoperative analgesia (0.2 mg/kg s.c. bupivacaine) and fluid support (1 mL i.p. saline). The surgical site was prepared by removing all hair and sterilizing with Betadine. Lubricant was applied to both eyes to avoid corneal drying. Mice were anesthetized (isoflurane, 3.5% induction, 0.5-1.0% maintenance) under aseptic conditions with temperature maintenance (37.0°C ± 0.5°C) via rectal probe. Once anesthesia depth was confirmed by lack of response to foot pinch, a midline ventral neck incision was made on the right side. The common carotid artery (CCA) and internal carotid artery (ICA) were isolated and temporarily ligated using 7-0 sutures. The distal external carotid artery (ECA) was permanently ligated. For Koizumi procedures, a weight-matched silicon coated filament (Doccol 602145 for <23 g, 602245 for 23-28 g, 602345 for >28g; Doccol, Sharon, MA) was inserted through an incision in the CCA superior to the point of ligation. For Longa procedures, the incision was instead made in the ECA proximal to the point of ligation for insertion of the weight-matched filament (Doccol 602123 for <23 g, 602223 for 23-28 g, 602323 for >28 g). In both protocols, the ICA was unligated and the filament was advanced ~9 mm into the ICA to occlude the origin of the middle cerebral artery. The filament was secured with a temporary suture distal to the insertion point. The neck incision was sutured, and animals were removed from anesthesia and allowed to recover in a heated chamber. To provide a clinical correlate for ischemic severity, intra-ischemic behavior was assessed just prior to re-anaesthetizing using an ordinal five-point scale as previously described [22]. Briefly, mice were allowed to move freely and then suspended by the tail to assess the degree of unilateral deficit exhibited, where 0 = no deficit, 1 = forelimb weakness and torso turning to the affected side when suspended by tail, 2 = spontaneous circling to the affected side, 3 = unable to bear weight on the affected side, and 4 = no spontaneous locomotor activity or barrel rolling. Mice with no deficits were excluded. Just prior to the end of the ischemic window, mice were re-anaesthetized and cerebral blood flow (CBF) was assessed by laser speckle contrast imaging (LSCI) before the neck incision was reopened. The filament was withdrawn to enable recanalization, and either the CCA (Koizumi) or ECA (Longa) was permanently ligated. The CCA was opened in the Longa surgical cohorts following filament removal. Mice recovered in a heated chamber before being returned to their home cage. Mice were excluded if subarachnoid hemorrhage was detected. Sham procedures were performed following the same approach as the Longa technique without incision into the ECA or insertion of a filament to confirm that anesthesia exposure or surgical trauma did not affect CBF integrity, locomotion, or immune cell recruitment (Figure S1).

Laser Speckle Contrast Imaging (LSCI)

To enable serial imaging, mice were anesthetized and the skin (~1 cm diameter) overlying the skull cap was removed. A layer of optically clear glue (Norland Optical Adhesive 81, Edmund Optics) was coated over the skull and cured with UV (5% UV, 10-15 seconds exposure, 5 cm height from skull) (CS 2020, ThorLabs). Skull covers were placed at least 24 hours prior to tMCAO procedure to minimize any possibility of procedure-induced vascular injury or inflammation at the time of surgery. LSCI (PeriCam PSI HR; Perimed) was conducted prior to and after UV glue coating to confirm the glue did not alter CBF patterning. LSCI was then captured at skull cover placement (baseline), immediately prior to filament withdrawal (ischemia), immediately after filament withdrawal (reperfusion), and immediately prior to tissue collection (endpoint). Ischemia and reperfusion were confirmed by LSCI, acquired immediately before and after filament withdrawal, as previously described [23]. Imaging was performed for >15 seconds and one representative frame was used for pixel quantification. CBF was represented by a color-gradated scale of arbitrary perfusion units (a.u.) according to the random motion detected at each pixel, with high random motion corresponding to healthy perfusion (red) and low random motion corresponding to hypoperfusion (blue). Images were generated as color-coded heatmaps of a.u. on a scale of 0 to 300. Matlab (24.2.0.2712019) was used to isolate the ischemic hemisphere from background signal according to the regional difference in perfusion units. Within the ischemic hemisphere, perfusion units were classified according to color (blue = 0-60, cyan = 61-120, green = 121-180, yellow = 181-240, red = 241-300 a.u.) and the number of pixels corresponding to each category was calculated. A ‘speckle score’ was calculated at baseline, ischemia, reperfusion, and endpoint using a weighted scale from 1 to 5 according to the average pixel color. Specifically, each color classification was assigned an ordinal value (red = 5, yellow = 4, green = 3, cyan = 2, blue = 1). The average value of all pixels within the ischemic hemisphere was then calculated to derive the speckle score. For instance, 5 corresponded to 100% red pixels and 1 corresponded to 100% blue pixels.

Behavioral Testing

Behavioral testing was performed at baseline and at 24 hours (24h) and 72 hours (72h) post-reperfusion. For corner testing, mice were placed in a 30° opaque corner with their whiskers contacting the walls and allowed to escape by rearing to either side [24]. Ten trials were conducted, and laterality index was calculated as the absolute difference between left and right turns divided by total turns. For open field testing, mice were allowed to explore an open field (42 cm x 42 cm x 42 cm) for 10 minutes [25]. Behavior was recorded and processed using Handbrake (1.7.3). Anatomic landmarks were tracked via DeepLabCut-based pose estimation (0.2.1.7) using an optimized configuration of the open-source Super Animal Top View Mouse network [26, 27]. RStudio (4.4.1) was used to calculate locomotor metrics and generate heatmaps.

Tissue Collection

Brains were collected at 24h or 72h post-tMCAO. Mice allocated to immunofluorescence imaging received a retro-orbital injection of 10% Tomato Lectin, Dylight 649 (Vector Laboratories) 15 minutes prior to collection to label reperfused vasculature. Mice were deeply anesthetized (isoflurane, 5%) and transcardially perfused with cold phosphate buffered saline (PBS) followed by cold 4% paraformaldehyde (PFA) if allocated to immunofluorescent imaging or PBS alone for infarct size quantification. Endogenous TdTomato signal confirmed retention of adherent PMNs in the vasculature following perfusion, as has been previously observed (Figure S2) [17, 28]. Brains were dissected, meninges removed, and sectioned into 1 mm coronal slices using an adult mouse brain slicing block (5325; Zivic Instruments).

Infarct Volume Quantification

For infarct size quantification, brain slices were stained with triphenyl tetrazolium chloride (TTC) for 15 minutes to distinguish infarcted (white) from healthy tissue (red) [29]. Seven consecutive coronal slices were imaged using a near-infrared fluorescent imager (Azure 500; Azure Biosystems). Infarct and hemorrhage volumes were quantified by manual tracing in Fiji ImageJ, with edema correction applied using the ipsilateral-to-contralateral hemisphere ratio.

Immunostaining and Confocal Imaging

Brain slices for immunofluorescence were post-fixed in 4% PFA for 24 hours and cryoprotected in 15% and 30% sucrose. Slices were washed in 0.3% Triton-X in PBS and blocked in 10% goat serum for 2 hours at room temperature, and stained overnight at 4°C with NeuN primary antibody (1:1000, ABN90P; Sigma-Aldrich), followed by incubation with goat anti-guinea pig 405 secondary antibody (1:1000, SAB4600230; Sigma-Aldrich) for 2 hours at room temperature. Samples were stored in 0.02% sodium azide in PBS prior to imaging. Fluorescent imaging was performed using an Airyscan two-photon laser-scanning confocal microscope (LSM 980; Zeiss). Coronal slices at bregma -1mm, corresponding to the region of maximal ischemia, were selected for analysis. Widefield scans of the ischemic hemisphere were acquired at 10x magnification (z-step: 10 μm, z-stack: 200-250 μm) and high-resolution z-stacks for cell counts and morphological analyses were acquired at 20x magnification (z-step: 0.48 μm, z-stack: 100-120 μm) in at least three cortical fields of view (FOV). Image stitching and reconstruction were performed in Fiji ImageJ.

Imaging Analysis

Cell analyses of MG/MΦ and PMNs captured in 20x FOVs were conducted using Imaris Single Full with Clear View (Imaris Bitplane v.10.2, Oxford Instruments) image analysis software. Images were pre-processed using the Gaussian Filter function to enhance cell signal relative to background. Cell and vascular reconstructions were created using the Machine Learning Segmentation Surface Creation tool with each source channel. Surfaces were filtered according to area and intensity to exclude non-cellular signal. Cell volume, sphericity, and distance to other surfaces were calculated automatically for each cell within the surface.

Statistics

All statistical analyses were conducted using GraphPad Prism (10.4.2). Input data for Prism was collated in Excel. For all tMCAO experiments except survival analysis, statistical significance was assessed using two-way analysis of variance (ANOVA) for ischemic duration and reperfusion status with Tukey’s multiple comparison post hoc tests as indicated between cohorts with common ischemic time or reperfusion success. Statistical significance for sham experiments was assessed using one-way ANOVA for timepoint with Tukey’s multiple comparison post hoc tests as indicated. Survival analysis was conducted using the log-rank Mantel-Cox test. All data is reported as the mean; *p < 0.05, **p < 0.01, ***p < 0.001. Results of all statistical analyses are reported in Table S1.

RESULTS

Ischemic duration and surgical approach alter infarct topology and intra-ischemic behavior

Surgical approach and the duration of ischemia employed are widely variable across preclinical studies, but the effects of these differences in ischemia and reperfusion dynamics on stroke outcomes have yet to be fully elucidated. To simultaneously compare tMCAO iterations with variable ischemia and reperfusion, the Longa and Koizumi methods of tMCAO were employed using multiple ischemic durations. These techniques differ in the identity of the vessel used for filament insertion and sacrificed following recanalization: the Longa protocol uses the ECA to result in complete reperfusion, while the Koizumi protocol uses the CCA to achieve incomplete reperfusion (Figure 1a–b). Within the Longa and Koizumi cohorts, the duration of ischemia was varied to 30, 60, and 90 minutes, corresponding to ischemic durations commonly used in preclinical studies. Through these tMCAO variations, a spectrum of six surgical conditions was established: 30-minute Longa (30min L), 60-minute Longa (60min L), 90-minute Longa (90min L), 30-minute Koizumi (30min K), 60-minute Koizumi (60min K), and 90-minute Koizumi (90min K). The effect of these surgical variants on infarct topology was first investigated according to TTC staining to delineate ischemic and healthy tissue (Figure 1c–d). Infarct volumes tended to increase with ischemic duration (Figure 1e). Additionally, the Koizumi protocol demonstrated larger infarct volumes specifically in the 90min cohort. Although the 90min L group did not exhibit significantly larger infarcts than surgical approach-matched groups of shorter ischemic duration, it did demonstrate a higher frequency of hemorrhagic transformation than the 30min L cohort (Figure 1f). However, a difference in hemorrhage volume was only detected between the 30min L and 60min L cohorts (Figure 1g). To link infarct pathology to the severity of the initial ischemic insult, intra-ischemic behavior was measured using an ordinal five-point scale. More severe intra-ischemic behavior was observed with increasing ischemic duration in the Koizumi cohorts, as evidenced by an increased prevalence of high behavior scores (Figure 1h). Intra-ischemic behavior was also worse in the 90min K group relative to the 90min L group, similar to the trends observed in infarct volume. Mortality within 72 hours post-tMCAO did not differ according to ischemic duration or surgical approach (Figure 1i).

Fig. 1. Increased ischemic duration is associated with more severe infarct pathology and intra-ischemic behavior.

The surgical approach was varied based on the location of filament insertion: insertion via the external carotid artery induced complete reperfusion (Longa) (a) whereas insertion via the common carotid artery induced incomplete reperfusion (Koizumi) (b). Surgical diagrams were created using BioRender, where the red area of the brain represents decreased perfusion, and the gray area denotes the region affected by initial ischemic insult. Ischemic duration was varied in each model (30min, 60min, or 90min) and infarct volume for each condition was quantified via triphenyl tetrazolium chloride (TTC) staining of seven serial tissue sections (c-e). Hemorrhagic transformation was also assessed according to the presence of gross blood on TTC (f), and hemorrhage volume was quantified using the same serial sections (g). Intra-ischemic behavior was evaluated using an ordinal 5-point scale, ranging in severity from 0 (no deficit) to 4 (no spontaneous locomotor activity or barrel rolling) (h). Mortality was also tracked over the entire study duration, up to 72 hours post-reperfusion (i). N = 7 per group for infarct topology, n = 37-50 per group for intraischemic behavior and survival analysis. Data expressed as % population (f, h-i) or mean (e,g). *P < 0.05, **P < 0.01. P values were calculated using two-way ANOVA tests for ischemic duration and surgical approach with Tukey’s multiple comparisons when indicated.

Incomplete reperfusion induces a sustained drop in cerebral blood flow

Given that the Longa and Koizumi protocols differ principally in whether they access the common carotid artery, we investigated whether these surgical approaches and varying ischemic durations influence spatiotemporal CBF patterning. Serial LSCI was performed at baseline, immediately before and after filament withdrawal, and at endpoint to create spatial maps of surface CBF. LSCI was employed as it enabled spatial imaging of the superficial cortex (~1 mm) while minimizing unnecessary anesthesia, which would alter the inflammatory and pathological environment. First, successful ischemia/reperfusion and infarct localization were confirmed for each surgical approach based on visual changes in perfusion relative to baseline (Figure 2a). Mirrored regions of interest (ROIs) were overlaid onto the ipsilateral and contralateral hemispheres to quantify CBF according to the scale of arbitrary perfusion units (a.u.), where higher values correspond to greater perfusion within the ROI (Figure S3a). CBF did not differ across experimental cohorts at baseline or during ischemia, confirming comparable ischemia induction between surgical approaches. However, the Longa protocol exhibited significantly greater CBF than ischemic duration-matched Koizumi cohorts at reperfusion and endpoint (Figure 2b, S3b–c). CBF was unaffected by ischemic duration in both the Longa and Koizumi cohorts at ischemia, reperfusion, and endpoint. Although these ROIs were useful to establish a straightforward comparison of perfusion between groups, this approach did not enable assessment of CBF across the entire ischemic hemisphere or more nuanced evaluation of the perfusion dynamics driving overall differences. To compare the composition of the entire ischemic hemisphere in terms of CBF, Matlab code was generated to quantify the proportion of pixels in the ischemic hemisphere corresponding to each color-gradated quintile on the scale of a.u. (blue = 0-60, cyan = 61-120, green = 121-180, yellow = 181-240, red = 241-300 a.u.). A significant decrease in the proportion of red pixels (which correspond to significant perfusion) and corresponding increases in yellow/green/cyan pixels (which correspond to hypoperfusion) were observed at reperfusion and endpoint in the Koizumi cohorts relative to their ischemia-matched Longa counterparts, indicating a sustained shift toward hyporeperfusion using the Koizumi protocol (Figure 2c, S3d, Table S1). However, pixel composition of the ischemic hemisphere did not differ in response to varied ischemic duration within either surgical approach. Pixel composition was then distilled into a quantitative summary metric, referred to as speckle score, to confirm the relevance of these perfusion dynamics in defining overall CBF across the entire ischemic hemisphere (Figure 2d, S3e). Speckle scores were derived by assigning each color quintile an ordinal value (red = 5, blue = 0) and calculating the mean value across the ischemic hemisphere. Speckle score during ischemia was significantly higher in the 60min K group compared to the 60min L group. However, these CBF dynamics underwent a shift following reperfusion, as evidenced by lower speckle scores and, therefore, less perfusion in the Koizumi cohorts than their Longa counterparts across all ischemic durations. Interestingly, this decrease in CBF observed in the Koizumi groups was sustained at endpoint only in the 90min K group, suggesting a potential interplay between ischemic duration and longitudinal CBF dynamics.

Fig. 2. Cerebral blood flow in the ischemic hemisphere is longitudinally affected by surgical approach.

Cerebral blood flow patterning was mapped across the entire brain in all models at baseline, immediately before (ischemia) and after (reperfusion) filament withdrawal, and at endpoint using laser speckle contrast imaging (a). Cerebral blood flow was measured according to arbitrary perfusion units (a.u.) and displayed as a color scale (blue = 0 a.u., red = 300 a.u.). Regions of interest were overlaid onto each hemisphere to quantify perfusion at each timepoint (b). ROI perfusion in the ipsilateral hemisphere is reported as fold change relative to the time-matched contralateral hemisphere measurement. Perfusion dynamics were also tracked across the entire ipsilateral hemisphere according to the proportion of total pixels corresponding to each color-coded quintile of the perfusion scale (c). Statistical significance of each color-coded quintile is denoted in the corresponding table for each timepoint (# = statistically significant); statistical values for each pairwise comparison are listed in the supplement. The gradient of perfusion across the ischemic hemisphere was simplified to an overall speckle score according to the average color classification of all pixels at each timepoint (blue = 0, red = 5) (d). N = 15 per group. Data expressed as fold change relative to the contralateral hemisphere (b), % pixels (c), or mean (d). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. P values were calculated using two-way ANOVA tests for ischemic duration and surgical approach with Tukey’s multiple comparisons when indicated.

Duration of ischemia differentially alters acute locomotor function in the setting of complete and incomplete reperfusion

Motor function is a primary measure of stroke severity and reflects multiple pathological features, including initial ischemia, infarct evolution, and longitudinal blood flow dynamics. Therefore, we evaluated how the variable perfusion dynamics observed between the Longa and Koizumi protocols related to acute functional outcomes. Corner testing was performed at baseline, 24 hours- (24h), and 72 hours- (72h) post-tMCAO to identify functional deficits according to side preference [30]. No differences in laterality were observed across groups (Figure S4a). However, corner testing assesses a single binary metric of locomotion that can be biased by the starting position or repeated exposure, so laterality may not be sensitive enough to detect nuanced motor differences. To enable more robust evaluation of post-stroke locomotion, open field testing was also performed at the same timepoints, and anatomic landmarks were tracked using DeepLabCut machine learning-based pose estimation (Figure S4b). Representative heatmaps were generated for all cohorts at baseline, 24h, and 72h and metrics of locomotion were derived from tracking of the ‘mouse center’ landmark (Figure 3a–b, S4c). Locomotion was decreased at 24h in the Koizumi groups with increasing ischemic duration, as evidenced by decreases in distance traveled, average speed, border-center transitions, and ‘percent time moving’ (Figure 3c–f). At 72h, the only parameter still affected by ischemic duration in the Koizumi groups was ‘percent time moving’. Although this response to ischemic duration was not observed using the Longa protocol at 24h, the Longa cohorts did exhibit a similar sensitivity at 72h to the severity of ischemia. Specifically, decreases in distance traveled, average speed, and, to a lesser extent, border-center transitions were evident at 72h with increasing ischemic duration. The extent of reperfusion had a limited effect on motor function, with lower ‘percent time moving’ only in the 30min L cohort relative to the 30min K group at 24h and 72h. Cumulatively, these data demonstrate that ischemic duration differentially affects locomotor function throughout acute stroke evolution in the Longa and Koizumi protocols. Further, this may implicate reperfusion-mediated elements of tMCAO, including hypoperfusion and reperfusion-induced inflammation, as time-dependent drivers of functional outcomes.

Fig. 3. Ischemic duration differentially mediates acute locomotor outcomes in the Longa and Koizumi models.

Representative heatmaps of open field activity were generated for each cohort at 24h (a) and 72h (b) to enable visualization of changes in overall locomotion. Corresponding representative baseline heatmaps are included in the supplement. Tracking data from the mouse center landmark were used to quantify distance traveled (c), average speed (d), border-center transitions (e), and percent time moving (f) at 24h and 72h. All open field metrics are reported as fold change relative to the subject’s baseline. N = 12-15 per group per timepoint. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. P values were calculated using two-way ANOVA tests for ischemic duration and surgical approach at each time point with Tukey’s multiple comparisons as appropriate.

Myeloid cell response is influenced by ischemic duration and surgical approach

These data point to a role for ischemic duration and surgical approach in mediating longitudinal stroke evolution. Given that neuroinflammation, specifically through the actions of myeloid cells, has been implicated in disease progression, we utilized transgenic reporter mice with markers for MG/MΦ (Csf1r-EGFP) and PMNs (Ly6G-TdTomato) to investigate how the Longa and Koizumi protocols at varying levels of ischemia influence the acute immune response. Although the Ly6G-TdTomato reporter is somewhat promiscuous and labels a subset of neurons in addition to PMNs, these cells exhibited spatial segregation: TdTomato+ neurons were confined to the striatum (and more prominent in the contralateral hemisphere) while PMNs localized to cortical/subcortical regions of the infarct. Therefore, histological analyses were restricted to the ischemic cortex/subcortex to avoid neuronal biasing. Further, TdTomato+ neurons and PMNs were distinguished by size and morphology, as neurons were up to 2-fold larger than PMNs and exhibited branched morphology. These differentiating metrics were validated via NeuN staining, which consistently localized to cells identified as neurons based on location and morphology (Figure S5).

Morphological and spatial features of MG/MΦ were evaluated by creating surface reconstructions using Imaris image analysis software to link ischemia and reperfusion variability with innate immune cell patterning (Figure S6). Representative images were selected by mean cell counts. At 24h, the number of MG/MΦ in the ipsilateral hemisphere did not vary between cohorts (Figure 4a–b). MG/MΦ morphology, as assessed by cell volume and sphericity, also did not differ between experimental groups at 24h (Figure 4c–d). However, the spatial association of MG/MΦ and PMNs was affected by ischemic duration, as evidenced by a decrease in MG/MΦ distance to the nearest PMN in the 90min L cohort relative to the 30min L group (Figure 4e). At 72h, MG/MΦ count was lower in the 30min K group than the 30min L group (Figure 4f–g). Differences in MG/MΦ volume and sphericity were also observed between the 30min and 60min groups based on reperfusion status (Figure 4h–i). Additionally, an increase in MG/MΦ sphericity was observed with increasing ischemic duration, but only in the Longa groups. Ischemic duration also mediated MG/MΦ proximity to PMNs at 72h, but only within the Koizumi cohorts, in contrast to the trend observed at 24h (Figure 4j).

Fig. 4. Microglia/macrophage cell recruitment and position are influenced by variability in model physiology.

Representative images of MG/MΦ (green) and vasculature (magenta) in the cortical region of the infarct at 24h post-tMCAO (a). MG/MΦ count per FOV (b), volume (c), sphericity (d), and position relative to the nearest PMN (e) were quantified at 24h using Imaris surface reconstruction. Representative images of MG/MΦ and vasculature were also captured for each experimental cohort at 72h (f). MG/MΦ count (g), volume (h), sphericity (i), and distance to nearest PMN (j) were measured using the same Imaris-based approach as was employed at 24h. N = 5 per condition per time point. Scale = 100 μm. Each data point represents an average of 2-3 technical replicates for a surgical subject. *P < 0.05, **P < 0.01. P values were determined using two-way ANOVA tests for ischemic duration and surgical approach at each timepoint with Tukey’s multiple comparisons as needed.

Similar morphological analyses were also performed for PMNs. Additionally, given that PMNs have been reported to further contribute to infarct evolution through microvascular obstructions when retained in the vasculature, the proportion of intravascular PMNs was also quantified [31, 32]. At 24h, PMN count increased between the 30min L and 60min L cohorts (Figure 5a–b). However, PMN volume, sphericity, and position relative to the vasculature did not significantly differ across cohorts (Figure 5c–e). Given this lack of differential response at 24h, we hypothesized that this timepoint may be too early to detect changes in PMN patterning mediated by ischemia and reperfusion dynamics, as PMN infiltration has been demonstrated to peak around 3 days post-stroke [33]. At 72h a significant increase in PMN count was observed with increasing ischemic duration regardless of protocol used, although this effect was more pronounced in the Koizumi cohorts (Figure 5f–g). PMN volume differed between Longa and Koizumi protocols, as evidenced by a decrease in the 60min K cohort relative to the 60min L group (Figure 5h). However, PMN sphericity and the proportion of PMNs retained intravascularly remained unchanged across surgical conditions (Figure 5i–j). Cumulatively, these data confirm unique effects of the Longa and Koizumi protocols, as well as ischemic duration, on the spatial and morphological profiles of myeloid cells during the acute stages of I/RI.

Fig. 5. Neutrophil infiltration and morphology are defined in part by tMCAO physiology.

Representative images of PMNs (red) and vasculature (magenta) in the ischemic cortex at 24h (a). At 24h, the number of PMNs (b), cell volume (c), sphericity (d), and % intravascular PMNs (e), were quantified via Imaris surface reconstruction. Representative images of PMNs and vasculature in the ischemic region of the cortex at 72h (f). PMN count per FOV (g), volume (h), sphericity (i), and % intravascular PMNs (j) were also measured at 72h. N = 3-5 per condition at each time point based whether PMNs were present in samples used for MG/MΦ analysis. Scale = 100 μm. Each data point represents an average of 2-3 technical replicates for a surgical subject. *P < 0.05, **P < 0.01, ***P < 0.001. P values were determined using two-way ANOVA tests for ischemic duration and surgical approach per timepoint with Tukey’s multiple comparisons if indicated.

DISCUSSION

Despite effective recanalization procedures, persistent functional deficits remain a challenge for post-stroke management [3]. I/RI-mediated inflammation contributes significantly to infarct evolution beyond the initial ischemic insult and drives functional outcomes [7]. In particular, myeloid cells, including PMNs and MG/MΦ, have been linked to stroke progression and are among the first immune cells to respond to ischemia, thereby likely mediating later stages of I/RI [31, 32, 34–41]. The tMCAO model is widely used to simulate stroke I/RI in preclinical studies aimed at identifying mechanisms and potential therapeutic targets involved in the accompanying immune response. However, the tMCAO model, performed through either the Longa or Koizumi surgical techniques, results in distinct outcomes, even though the two protocols are used interchangeably across a wide range of ischemic durations. The Longa and Koizumi approaches differ in the identity of the vessel used for filament insertion, with preservation of the CCA being a characteristic feature of the Longa approach, facilitating successful reperfusion as would be seen in the majority of patients clinically. Accordingly, the Longa approach is likely more applicable to studying immune-cell responses, cerebroprotective avenues, and brain resilience in the modern thrombectomy era. On the other hand, while the Koizumi approach requires direct CCA involvement and prohibits complete reperfusion, the result more closely mimics an ineffective reperfusion observed in a subset of patients. As a result of the heterogeneity between models, metrics of infarct severity and functional outcome, as well as the contribution of I/RI, have been highly discordant across studies.

Although the middle cerebral artery is occluded in both tMCAO protocols, the volume of infarcted tissue and localization of the lesion vary dramatically. While most studies observe striatal injury following ischemia, there is a lack of consensus as to the involvement of the cortex, as well as structures not supplied by the MCA, including the hippocampus, thalamus, and amygdala [11, 42, 43]. Accordingly, the severity of functional deficits is highly variable, with conflicting reports of deficits or improvements in behavioral outcomes, as well as locomotor activity that is indistinguishable from that of animals undergoing sham surgery using the same motor function metrics [44–46]. These discrepancies in measures of injury severity also extend to the inflammatory response in terms of magnitude, timing, and cellular constituents. A meta-analysis examining post-stroke immune cell infiltration found high variability across studies in the density of MG/MΦ and PMNs in infarcted tissue throughout the acute phase of infarct evolution [9]. Further, the timing of maximal PMN influx remains unclear, with some reports citing a peak as early as 12 hours post-tMCAO and others observing higher levels of trafficking 2-3 days after surgery [33, 35, 47]. The interpretation of these contradictory results is complicated likely because of an inconsistency in the tMCAO methodology employed across studies, calling into question how the surgical approach and duration of ischemia utilized affected the reported outcomes.

In response to the confounding variables introduced by surgical variation, other groups have performed direct comparisons between the Longa and Koizumi protocols, but these analyses have primarily been limited to imaging- and histology-based quantification of infarct volume and CBF [48, 49]. In separate cohorts, the effect of ischemic duration on infarct topology has similarly been evaluated [34, 43, 50, 51]. Some studies have also sought to correlate lesion volume with functional outcomes using neurological deficit scores such as the Zea Longa 5-point score and modified neurological severity scores [52]. However, a comprehensive assessment linking infarct severity, neuroinflammation, and motor function in response to tMCAO protocol or ischemic duration has yet to be performed. Comparison of the immune response across surgical conditions has been particularly limited, with few studies linking injury severity to inflammatory metrics, such as cytokine levels and PMN-vascular associations [53, 54]. To our knowledge, the effect of these surgical conditions on innate immune cell recruitment has not been investigated, despite previous findings highlighting the responsiveness of myeloid cell accumulation to occlusion location and transience [9, 55]. Further, although the collective injury burden of ischemia and reperfusion necessitates that these facets of the tMCAO model be considered in parallel, the consequences of the Longa and Koizumi protocols have never been considered in relation to ischemic duration. This comparison is critical to uncouple the effects of ischemic burden and reperfusion status from other surgical variables that may differ between groups.

To address these concerns and the inconsistency of outcomes, this preclinical study was designed to examine how variability in ischemic duration and the route of reperfusion affects the myeloid cell response and subsequent pathological and functional outcomes. This work was performed according to the standard methodology for both the Longa and Koizumi procedures, which employ identical surgical conditions except for the length of the filament used and the route of filament insertion (as are typically distinct between models). In keeping all other elements of the surgical procedure consistent across cohorts, our approach ensured that only ischemic duration and reperfusion features drove differences in our findings, which was confirmed by a lack of difference in ischemic severity across protocols. We employed serial perfusion imaging, behavioral testing, and confocal imaging of transgenic myeloid cell reporters to establish a comprehensive comparison of the Longa and Koizumi tMCAO protocols at varying ischemic durations.

First, our results indicate that infarct severity, post-stroke locomotion, and myeloid cell features are more robustly mediated by ischemic duration than by the difference in reperfusion achieved by the Longa and Koizumi protocols. Increasing ischemic duration resulted in larger infarcts, increased risk of intraparenchymal hemorrhage, more severe behavioral deficits, and more robust immune cell recruitment. While there were differences in infarct volume, functional outcomes, and immune response between the Longa and Koizumi protocols, many of these effects were restricted to a single ischemic duration. These ischemia-specific responses between protocols highlight an important interplay between ischemia and reperfusion, where sensitivity to hypoperfusion, as is characteristic of the Koizumi protocol, is intrinsically linked to the severity of the initial ischemic insult. Therefore, differences in the immune response may be due to increased damage-associated molecular pattern or cytokine release induced by the initial ischemic event. In this scenario the subsequent reperfusion would be a necessary therapy that also promotes the opening of vascular channels, thereby fueling further inflammatory cell recruitment [56–58].

Further emphasizing the interrelatedness of ischemia and reperfusion dynamics, this study revealed that ischemic duration differentially mediated motor outcomes according to the degree of reperfusion achieved in the Longa and Koizumi protocols. In the Koizumi protocol, increasing ischemic duration corresponded to more severe behavioral deficits at 24h, but apart from ‘percent time moving’, this effect was not sustained at 72h. In contrast, animals subjected to the Longa protocol showed a similar pattern of ischemic duration correlating to worse behavioral outcomes, but this effect was only apparent at 72h rather than 24h. This temporal difference in functional metrics across surgical conditions points to unique facets of tMCAO pathology that influence behavior over time: hyporeperfusion induces an immediate exacerbation of the initial insult, while complete reperfusion enhances I/RI, worsening outcomes at later time points.

This study also revealed that ischemic duration and reperfusion status differentially alter the profile of MG/MΦ and PMNs at 72h post-tMCAO. PMN density alone, rather than PMN morphology or position, was more robustly affected by the time of ischemia, suggesting that the magnitude of the PMN response depends on the severity of the initial ischemic event. MG/MΦ density and morphology at 72h were also sensitive to variable surgical conditions but, in contrast to PMNs, these differences were primarily between the Longa and Koizumi protocols. The unique sensitivities of PMNs and MG/MΦ to ischemia and reperfusion elements of the protocols that were observed align with the differential roles that these myeloid cell populations may play in I/RI. As PMNs are peripheral immune cells recruited to the site of ischemia to exert transient pro-inflammatory effects, PMN infiltration appears to be triggered by the initial ischemic event, with lesser effects observed during infarct evolution after the initial wave of cell activation [33]. In contrast, MG/MΦ respond rapidly to the inciting injury and persist in the ischemic tissue to contribute longitudinally to inflammation and injury resolution. Therefore, the evolving response of MG/MΦ to differences in reperfusion, rather than their initial response to the ischemic insult, is captured at the acute timepoints studied [59, 60]. These differential effects confirm that myeloid cells uniquely respond to ischemic duration and surgical approach in tMCAO, guiding model selection for future preclinical studies investigating the contributions of these immune cell populations.

In addition to these individual responses, the position of MG/MΦ relative to PMNs was also influenced by ischemic duration. MG/MΦ and PMNs participate in cell-cell interactions in neuroinflammation, so proximity was measured as a surrogate for cellular association, with shorter distances suggestive of greater potential for interaction. The spatial association of MG/MΦ and PMNs in response to increasing ischemic duration is consistent with the ischemia-mediated increase in PMN density, implicating PMNs as the likely drivers of this relationship. Notably, our immunofluorescent confocal imaging revealed that a subset of MG/MΦ completely overlapped with PMNs, pointing to contact-mediated interactions between these populations. Accordingly, it may be worthwhile to further investigate the changing relationship between MG/MΦ and PMNs in each of the tMCAO iterations studied to define how these interactions shape the acute tissue environment post-stroke.

This study has several strengths, including the application of multiple tMCAO protocols, non-invasive live-animal imaging, and transgenic fluorescent reporters. However, there are limitations as well. Analysis of myeloid cells did not include evaluation of cell function. Future studies will aim to link immune cell function with position and morphology, as well as extend these analyses to subacute and chronic timepoints, such as days 7- and 28-post stroke, to evaluate the longitudinal effects of ischemic duration and surgical approach on neuroinflammation and stroke pathology. This study also utilized the Csf1r-EGFP reporter to label both MG and MΦ in parallel with the Ly6G-TdTomato reporter as a means to investigate the spatial relationships between myeloid cell populations. Additional studies should be performed using the Cx3cr1-Cre-YFP-TdTomato lineage-tracing strain to differentiate the contributions of MG and MΦ in the absence of PMN labeling. It is also important to note that although the Longa and Koizumi models are both used to recapitulate stroke I/RI, the two tMCAO approaches differ not just in surgical procedures, but also in reperfusion dynamics and pathophysiological responses. The results we report here are informative of the overall effects of each selected model.

Overall, these findings highlight how ischemic duration and surgical protocol interactively shape infarct severity, acute behavior, and the myeloid cell response in tMCAO, with effects that are temporally distinct depending on the degree of reperfusion achieved. This methodological framework for selecting tMCAO conditions that best model specific neuroinflammatory and behavioral patterns may improve the design of future preclinical studies and the translational relevance of stroke research.

Supplementary Material

1

Fig. S1 Sham tMCAO surgery does not affect cerebral blood flow, locomotion, or acute myeloid cell recruitment Cerebral blood flow was mapped across the entire brain surface at timepoints corresponding to those measured in the tMCAO models (a). Cerebral blood flow was measured according to arbitrary perfusion units (a.u.) and displayed as a color scale (blue = 0 a.u., red = 300 a.u.). Normal perfusion was confirmed across all timepoints according to ROI perfusion fold change in the ipsilateral hemisphere relative to the contralateral hemisphere (b), the proportion of total pixels corresponding to each color-coded quintile of the perfusion scale (c), and overall speckle score (d). Locomotion was measured at baseline, 24h, and 72h post-surgery according to average moving speed (cm/s) using tracking data from the mouse center landmark (e). Representative images of MG/MΦ (green), PMNs (red), and vasculature (magenta) in the ipsilateral cortex at 24h (left) and 72h (right) (f). The number of MG/MΦ was quantified by manual cell counts of Iba1 staining normalized to the average number of Iba1+ cells in control samples (g). No PMNs were detected at any timepoint, so quantifications were excluded. Scale = 50 μm (f). N = 3-6 for all metrics. P values were determined using one-way ANOVA tests for timepoint with Tukey’s multiple comparisons if indicated.

2

Fig. S2 Adherent neutrophils were retained in the vascular lumen following transcardial perfusion Representative high-resolution imaging captured throughout the ischemic hemisphere enabled visualization of intraluminal PMNs (denoted by blue asterisks) (a). PMN position within the vessel was confirmed according to PMN/vessel signal overlap across all orthogonal planes (b). The ratio of PMN/vessel volume overlap was quantified for each PMN across all surgical cohorts at 24h (c) and 72h (d), where volume overlap ratio >0.9 corresponded to intraluminal PMNs (blue asterisk). Scale bars = 25 μm.

3

Fig. S3 Cerebral blood flow patterning differed between Longa and Koizumi models. Corresponding ROIs were overlaid onto the ipsilateral and contralateral hemispheres across all timepoints for each surgical subject (a). Perfusion was quantified at baseline according to the average a.u. value within the ipsilateral ROI normalized to the contralateral ROI to confirm baseline perfusion did not differ across models (b). Reperfusion was confirmed according to ipsilateral ROI perfusion at reperfusion relative to ROI-matched perfusion at baseline (c). Pixel composition was also mapped across the ipsilateral hemisphere at baseline (d), and a baseline speckle score was assigned to each subject based on mean pixel classification (red = 5, yellow = 4, green = 3, cyan = 2, blue = 1) (e). N = 15 per group. Data expressed as fold change relative to the contralateral hemisphere (b), % perfusion relative to the ipsilateral hemisphere at baseline (c), % pixels (d), or mean (e). P values were calculated using two-way ANOVA tests for ischemic duration and surgical approach with Tukey’s multiple comparisons when indicated.

4

Fig. S4 Overall locomotion was evaluated according to corner and open field testing Unilateral motor deficits were evaluated at 24h and 72h according to laterality index, which was calculated using a 10 turn corner testing paradigm (a). Anatomic landmarks were tracked during open field testing using DeepLabCut pose estimation, where each colored label corresponds to a point used for tracking (b). Representative heatmaps of locomotion during open field testing were generated for each surgical condition at baseline (c). Data expressed as difference in laterality relative to the subject’s pre-surgical baseline (a). P values were calculated using two-way ANOVA tests for ischemic duration and surgical approach at each timepoint.

5

Fig. S5 TdT+ neurons and neutrophils were differentiated by NeuN staining in addition to location and morphology Widefield imaging of TdTomato and NeuN fluorescence throughout the brain (a) demonstrated that TdT+NeuN+ neurons (denoted by white arrows) localize to deeper striatal regions typically in the contralateral hemisphere (b), while TdT+NeuN- neutrophils (denoted by yellow arrows) are primarily restricted to the cortex and subcortex of the ipsilateral hemisphere (c-d). Infarcted tissue is denoted by dashed blue line; white boxes correspond to regions of 20x imaging in (b-d). Scale bars = 500 μm (a) and 100 μm (b-d).

6

Fig. S6 Imaris surface reconstruction enabled quantification of microglia/macrophage and neutrophil morphology and spatial relationships 20x confocal images of MG/MΦ (green), PMNs (red), and vasculature (magenta) captured in the ipsilateral cortex/subcortex were used as input (a) for Imaris image analysis software-based surface reconstruction (b). Sphericity, volume, distance to the nearest immune cell, and distance to the nearest vessel were automatically calculated from the reconstruction for each MG/MΦ (c) and PMN (d) in a sample. For each cell feature, the average of all cells quantified for a biological replicate was reported. MG/MΦ associated with PMNs denoted by white asterisks, MG/MΦ associated with vessels denoted by white arrows, PMNs associated with vessels denoted by blue arrows. Scale bars = 100 μm (a) and 30 μm (b).

7

Table S1 Statistical analyses for all presented data

• Ischemic duration more robustly drives stroke outcomes than surgical approach

• Motor outcomes diverge temporally by ischemic duration-surgical approach interplay

• Myeloid cells exhibit unique cell type-specific responses to model variables

ABBREVIATIONS

30min K

30-minute Koizumi model

60min K

60-minute Koizumi model

90min K

90-minute Koizumi model

30min L

30-minute Longa model

60min L

60-minute Longa model

90min L

90-minute Longa model

CBF

Cerebral blood flow

CCA

Common carotid artery

24h

24 hours post-reperfusion

72h

72 hours post-reperfusion

ECA

External carotid artery

ICA

Internal carotid artery

I/RI

Ischemia/reperfusion injury

LSCI

Laser speckle contrast imaging

LVO

Large vessel occlusion

MG

Microglia

MΦ

Bone marrow-derived macrophages

PMN

Polymorphonuclear neutrophils

tMCAO

Transient middle cerebral artery occlusion

TTC

Triphenyl tetrazolium chloride

Data Availability:

The datasets used and/or analyzed in the current study are available from the corresponding author on reasonable request.