Abstract graphic file with name cn3c00826_0005.jpg Ischemic strokes, prevalence and impactful, underscore the necessity of advanced research models closely resembling human physiology. Our study utilizes nonhuman primates (NHPs) to provide a detailed exploration of ischemic stroke, integrating neuroimaging data, behavioral outcomes, and serum proteomics to elucidate the complex interplay of factors involved in stroke pathophysiology. We observed a consistent pattern in infarct volume, peaking at 1-month postmiddle cerebral artery occlusion (MCAO) and then stabilized. This pattern was strongly correlated to notable changes in motor function and working memory performance. Using diffusion tensor imaging (DTI), we detected significant alterations in fractional anisotropy (FA) and mean diffusivity (MD) values, signaling microstructural changes in the brain. These alterations closely correlated with the neurological and cognitive deficits that we observed, highlighting the sensitivity of DTI metrics in stroke assessment. Behaviorally, the monkeys exhibited a reliance on their unaffected limb for compensatory movements, a common response to stroke impairment. This adaptation, along with consistent DTI findings, suggests a significant impact of stroke on motor function and spatial perception. Proteomic analysis through MS/MS functional enrichment identified two distinct groups of proteins with significant changes post-MCAO. Notably, MMP9, THBS1, MB, PFN1, and YWHAZ were identified as potential biomarkers and therapeutic targets for ischemic stroke. Our results underscore the complex nature of stroke and advocate for an integrated approach, combining neuroimaging, behavioral studies, and proteomics, for advancing our understanding and treatment of this condition. Keywords: stroke, magnetic resonance imaging, nonhuman primate model, proteomics Introduction Stroke poses a major global health challenge, standing as the primary cause of long-term disability and the second leading cause of death worldwide. Ischemic strokes, accounting for about 77% of cases, arise from reduced cerebral blood flow due to arterial blockages, leading to tissue infarction and extensive cellular damage.^[44]1 The consequences of stroke are significant, manifesting as neurological impairments like contralateral hemiparesis, which profoundly affect daily life. The severity and type of these impairments depend on the affected brain areas, often resulting in motor deficits^[45]2,[46]3 that impacts over 80% of stroke survivors, particularly in upper limb movement, drastically impacting their quality of life.^[47]4 Animal models, especially nonhuman primates (NHPs), play a crucial role in stroke research for developing effective treatments. While rodent models have contributed significantly, their physiological and anatomical disparities from humans limit the translating potential of research findings.^[48]5 NHP models, due to their closer genetic and physiological resemblance to humans, offer a more accurate representation of human stroke pathology and recovery mechanisms.^[49]6−[50]8 These models are particularly valuable for long-term neuroimaging and behavioral assessments, providing a detailed view of stroke effects over time.^[51]9 Neuroimaging advancements, especially diffusion tensor imaging (DTI), have significantly improved our understanding of stroke. DTI allows in vivo assessment of tissue integrity and connectivity, providing potential biomarkers for disease progression.^[52]10,[53]11 It has been effective in tracking changes in white matter and gray matter during stroke in both rodent^[54]12 and human^[55]13 studies. DTI metrics, such as fractional anisotropy (FA) and mean diffusivity (MD), provide insights into the microstructural integrity of neural tissue,^[56]14,[57]15 with lower FA values linked to poorer motor outcomes in stroke patients.^[58]16 Chronic white matter injury manifests as increased MD and decreased FA have been consistently observed in stroke patients^[59]17 and NHP stroke model.^[60]18 Proteomics has become an indispensable tool in stroke research, identifying biomarkers and therapeutic targets while illuminating proteomic alterations across different stroke phases.^[61]19 Extensive studies in rodent and NHP models have uncovered over 7000 differentially expressed proteins in ischemic core and penumbra regions, implicated in critical processes like cell death, survival, and recovery.^[62]20 Serum proteomics in NHP stroke models provides insights into cellular and biochemical pathways involved in stroke progression from acute to chronic phases in rodents^[63]21 and NHPs.^[64]22 This study aimed to explore the intricate stages of stroke in NHP stroke models by examining changes in brain microstructural, pathological indices, neurological alterations, and proteomic changes. We hypothesize that the disruption of blood supply, coupled with neuronal cell damage and brain matter integrity impairment, leads to cerebral edema, dysfunction, and structural abnormalities. Furthermore, we anticipate that motor injury symptoms are associated with alterations in DTI metrics, necessitating a thorough investigation of the correlations between DTI parameters and neurological features. Through serum proteomic mapping, we aim to explore the parallels between NHP stroke models and human stroke, focusing particularly on biomarkers reflective of brain microstructural integrity and their association with the pathophysiological mechanisms of stroke. Results Our study employed middle cerebral artery occlusion (MCAO) using bipolar electrocoagulation to generate ischemic stroke models in NHPs. The integration of neuroimaging data, behavioral assessments, and detailed serum proteomic analysis were conducted at both 1-month and 3-month post-MCAO. Our goal was to unravel the intricate interplay among various factors influencing stroke pathophysiology, providing a more holistic understanding of the condition. Changes in Infarct Volume and Microstructure in the Brain Longitudinal monitoring of the infarct volume using T2-weighted images showed maximal ischemic lesion expansion at 1-month post-MCAO (p = 0.0146), as shown in [65]Figure [66]1 A. The total infarct volume remained consistent between 1-month and 3-month post-MCAO (p = 0.9554), suggesting no significant reduction of lesion size during this period. The infarct area, marked in red on each axial MRI image, varied in location across the subjects. This variation is depicted in [67]Figure [68]1A, which aligns MRI findings with Paxinos Macaque Atlas.^[69]23 The location of infarct areas varied across several critical regions, including the orbitofrontal cortex, insular cortex, and basal ganglia components, such as the putamen, caudate nucleus, olfactory tubercle, and globus pallidus. These regions are pivotal for motor control, cognitive processes, and sensory integration. An exceptionally high intraclass correlation coefficient (0.998) between two independent readers confirmed the reliability of these volume measurements. Figure 1. [70]Figure 1 [71]Open in a new tab MRI image analysis. (A) T2-weighted images display the ischemic lesion (marked in red) against macaque brain atlas sections. The infarct area expands notably at 1-month post-MCAO, followed by a noticeable reduction by 3-month. (B) In the lesioned left hemisphere, ipsilateral fractional anisotropy (FA) values decrease at 1-month post-MCAO and partially recover by 3-month but remain reduced compared to the contralateral side. Mean diffusivity (MD) values show an increasing trend over the 3-month period. (C) This segment contrasts FA and MD values between ipsilateral and contralateral regions across six ROIs: caudate nucleus (Cd), olfactory tubercle (OT), putamen (Pu), globus pallidus (GP), orbitofrontal cortex (OC), and insular cortex (IC), measured at baseline, 1-month, and 3-month post-MCAO. Statistical significance levels are indicated as *** (p < 0.001), ** (p < 0.01), and * (p < 0.05). Data are presented as mean ± SEM, with’Con’ for contralateral and’IP’ for ipsilateral. DTI metric and FA and MD values were used to evaluate ischemic-induced microstructure alterations in the brain, as shown in [72]Figure [73]1 B. FA values provide information about the integrity and directionality of matter tracts in the brain, and MD values measure the average rate of water diffusion within a tissue, regardless of its direction. In general, reduced FA and increased MD values are often observed in areas of brain pathology. A significant FA decrease occurred at 1-month post-MCAO compared to baseline, indicating disrupted matter integrity, which partially recovered by 3-month post-MCAO. Across all ipsilateral ROIs, FA values were significantly lower than those on the contralateral side (p < 0.05), excepting in the olfactory tubercle. Notable changes in FA were observed in the caudate nucleus, putamen, globus pallidus, and insular cortex, with significant differences between baseline and the 3-month post-MCAO, as shown in [74]Figure [75]1C. In contrast, MD values consistently increase from baseline to 3-month post-MCAO, suggesting increased water content and breakdown of normal cellular structures. At 3-month post-MCAO, significant increases in MD were shown in the caudate nucleus, putamen, orbitofrontal cortex, and insular cortex compared to the baseline ([76]Figure [77]1C). Additional DTI metrics, axial diffusivity (AD) and radial diffusivity (RD), further elucidated changes in axonal integrity and myelination status within the brain, detailed in the [78]Supporting Information Figure S1. Behavioral Outcomes Behavioral assessments before MCAO, and at 1-month and 3-month post-MCAO, encompassed evaluations of neurologic deficit evaluation, motor function analysis (hill and valley staircase task, HVST), working memory assessment (delayed-response task, DRT), and cognitive function evaluation (object-retrieval detour task, ORDT). Initially, the monkeys exhibited normal neurologic functions with an average score of 6.00 ± 1.63. However, significant neurological deficits, including right limb paralysis and reduced tone, were observed at 1-month post-MCAO (average scores: 29.67 ± 0.94, p < 0.0001), with minimal recovery by 3-month post-MCAO (average score: 28.33 ± 0.94, p = 0.2302), as shown in [79]Figure [80]2 A. [81]Figure [82]2B showed that all monkeys nearly completed the HVSTs using both upper limbs before MCAO. At 1-month and 3-month post-MCAO, impaired right upper limb function hindered task performance, while left limb function remained unaffected. The DRT results, shown in [83]Figure [84]2C, indicate an initial average correct response rate of 85.00 ± 17.80, which significantly dropped to 63.33 ± 22.11 at 1-month post-MCAO (p = 0.008), suggesting a decline in working memory, with no further deterioration at 3-month (61.67 ± 17.95, p = 0.8283). The ORDT results showed a success rate of 91.04 ± 6.92% before MCAO, with barrier hits of 1.73 ± 1.69, as shown in [85]Figure [86]2D. A significant decrease in success rate (81.04 ± 10.97%, p = 0.0075) and an increase in barrier hits (4.20 ± 4.78, p = 0.0793) were observed at 1-month post-MCAO, indicating cognitive impairment. By 3-month MCAO, both success rate (93.75 ± 4.11, p = 0.0004) and barrier hits (0.67 ± 1.07, p = 0.0116) significantly improved, closely with pre-MCAO performance. Figure 2. Figure 2 [87]Open in a new tab Behavioral assessment. (A) Neurologic score showed a significant increase at 1-month and 3-month post-MCAO compared to the baseline. (B) Motor function in the left upper limb remained unaffected in both hill and valley staircase tasks (HVST) post-MCAO, maintaining baseline levels. Conversely, right upper limb was notably impaired post-MCAO, reflecting a decline from baseline performance. (C) In the delayed-response task (DRT), a significant decrease in the average correct response rate was observed at both at 1-month and 3-month post-MCAO. (D) The object retrieval detour task (ORDT) showed dynamic changes post-MCAO: the success rate of first-attempt reward retrieval significantly decreased at 1-month but recovered by 3-month, while the frequency of barrier hits increased initially at 1-month post-MCAO, indicating increased attempts to reach the reward through the transparent wall and then showed marked improvement by 3-month. Data present mean ± SEM of from 3 monkeys across 5 independent experiments. Significance denoted as *** (p < 0.001), ** (p < 0.01), and * (p < 0.05) versus baseline and ### (p < 0.001) versus 1-month, all determined by Student’s t test. Serum Protein Identification and Quantitation Using liquid chromatography–mass spectrometry (LC–MS/MS), we analyzed the tryptic peptides from serum protein samples collected before MCAO and at 1-month and 3-month post-MCAO. We identified a total of 5214 peptides and 799 proteins. Proteins exhibiting a fold change greater than 1.5 and a p value less than 0.05, as determined by Student’s t test, were considered differentially expressed. At 1-month post-MCAO, 60 proteins showed significant expression changes, increasing to 91 proteins at 3-month port-MCAO, as shown in [88]Figure [89]3 A. Hierarchical cluster analysis was performed to visualize these expression changes (details in [90]Figure S2), and using the Mfuzz package, six distinct clusters were identified based on differential expression criteria (fold change >1.5, p-value <0.05). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were conducted to delve into the biological significance of these clusters, as shown in [91]Figure [92]3B. By applying a more stringent fold change threshold of greater than 5.0 and p-value less than 0.05, 42 candidate proteins across cluster were identified (details in [93]Table S1), with notable upregulation in cluster 1 and cluster 2. Significant proteins involved in cell morphogenesis and migration included actin beta-like 2 (ACTBL2), actin gamma 1 (ACTG1), f-actin-capping protein subunit alpha-2 (CAPZA2), profilin 1 (PFN1), tropomyosin 4 (TPM4), moesin (MSN), WD repeat domain 1 (WDR1), S100 calcium binding protein A4 (S100A4), monoglyceride lipase (MGLL), coactosin-like f-actin binding protein 1 (COTL1), Ras suppressor protein 1 (RSU1), parvin beta (PARVB), and transgelin (TAGLN2). Proteins like matrix metallopeptidase 9 (MMP9), thrombospondin 1 (THBS1), and heat shock protein HSP 90-alpha (HSP90AA1) in cluster 1 and l-lactate dehydrogenase A chain (LDHA), actinin alpha 1 (ACTN1), coronin 1A (CORO1A), coronin 1B (CORO1B), glutathione S-transferase pi 1 (GSTP1), and tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein zeta (YWHAZ) in cluster 2 were identified as key in apoptotic regulation and were also detected. Myoglobin (MB) in cluster 1 and serpin family B member 1 (SERPINB1), associated with coagulation regulation, was also detected. Myoglobin (MB) in cluster 1 and serpin family B member 1 (SERPINB1) was also noteworthy. Utilizing the STRING database, we constructed a protein–protein interaction network for proteins in clusters 1 and 2, as illustrated in [94]Figure [95]3C. This approach enabled us to reveal potential interconnections and explore the broader functional implications of these proteins in the context of ischemic stroke pathophysiology. Figure 3. [96]Figure 3 [97]Open in a new tab Serum protein expression analysis. (A) Comparative graph depicting the count of differentially expressed proteins before MCAO and at 1-month and 3-month post-MCAO, highlighting temporal protein expression changes. (B) Application of the Mfuzz package for hierarchical cluster analysis identified 6 distinct clusters based on a fold change >1.5 and p-value <0.05. The left side is the line graphs of protein expression levels, and the vertical axis represents the relative expression levels of proteins, and the color of the line indicates the degree of membership of the protein in the current cluster. The right side shows a heatmap of expression levels, and the color map indicates the relative expression levels of proteins in the samples. (C) Visualization of the protein–protein interaction network for proteins in cluster 1 and cluster 2, utilizing the STRING database to elucidate potential interconnections and functional relationships. Correlations Between MRI Data, Behavioral Outcomes, and Serum Proteomics Correlation analysis between MRI data, behavioral outcomes, and serum proteomics were conducted, as shown in [98]Figure [99]4 . The relationships between infarct volume and behavioral outcomes were illustrated in [100]Figure [101]4A, revealing a significant positive correlation between infarct volume and neurological scores (p = 0.0004). Conversely, an inverse correlation was observed between infarct volume and accuracy in the DRT correct responses (p = 0.0123), while no significant correlation was found with the ORDT scores. These findings indicate that infarct volume is associated with alterations in neurological and working memory functions. [102]Figure [103]4B shows the correlations between DTI metrics and behavioral outcomes. FA values positively correlated with DRT correct responses (p = 0.0092), while negative correlations were observed with MD (p = 0.0270), AD (p = 0.0459), and RD (p = 0.0105) values. Neurological score negatively correlated with FA (p = 0.0024), but no significant correlations were identified with ORDT outcomes. Additionally, cluster 1 and cluster 2 negatively correlated with FA (p = 0.0433) and positively with MD (cluster 1: p = 0.0061, cluster 2: p = 0.0061), AD (cluster 1: p = 0.0045, cluster 2: p = 0.0311), and RD (cluster 1: p = 0.0045, cluster 2: p = 0.0031), as shown in [104]Figure [105]4C. [106]Figure [107]4D showed that both clusters negatively correlated with DRT correct responses (cluster 1: p = 0.0165, cluster 2: p = 0.0035) and positively correlated with neurological function scores (cluster 1: p = 0.0118, cluster 2: p = 0.0160)), indicating more severe deficits associated with higher cluster values. In addition, correlations between DTI metrics and serum protein are shown in [108]Table [109]1. FA values negatively correlated with proteins in cluster 1, including MMP9, THBS1, and MB, and MD values showed a positive correlation with proteins in cluster 2, including SERPINB1, ACTBL2, ACTG1, RUS1, MGLL, PFN1, WDR1, TAGLN2, YWHAZ, and ACTN1. Figure 4. [110]Figure 4 [111]Open in a new tab Correlations between infarct volume, DTI metrics, behavioral outcomes, and serum proteomics. (A) Negative correlation between infarct volume and delayed-response task (DRT) correct. A positive correlation with neurological score, and no significant correlations with object retrieval detour task (ORDT) success (p = 0.5422) or barrier hits (p = 0.9545). (B) Detail the correlations between DTI metrics, including fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radical diffusivity (RD), and behavioral outcomes: FA shows a significant positive correlation with both DRT correct and neurological scores (p = 0.0024), while MD, AD, and RD exhibit negative correlations with DRT correct. In contrast, no significant correlations are observed between any of the DTI metrics and ORDT success (FA, p = 0.1080, MD, p = 0.8432, AD, p = 0.7756, RD, p = 0.8432) and ORDT barrier hits (FA, p = 0.4121, MD, p = 0.2922, AD, p = 0.2502, RD, p = 0.2502). (C) Explore the correlation between serum proteomic clusters (cluster 1 and cluster 2) and DTI metrics, with both clusters showing significant correlations with each DTI metric. (D) Cluster 1 is negatively correlated with DRT and positively correlated with neurological scores. Cluster 2 exhibits a similar pattern, showing a negative correlation with DRT accuracy and a positive correlation with neurological scores. Significance levels are indicated as *** (p < 0.001), ** (p < 0.01), and * (p < 0.05). Table 1. Correlation Between DTI Metrics and Serum Protein[112]^a. FA __________________________________________________________________ MD __________________________________________________________________ AD __________________________________________________________________ RD __________________________________________________________________ protein r p-value r p-value r p-value r p-value cluster 1 MMP9 –0.9167 0.0013** 0.5333 0.1475 0.4 0.2912 0.6167 0.0857 THBS1 –0.7 0.0433* 0.85 0.0061** 0.7667 0.0214* 0.8667 0.0045** MB –0.9333 0.0007*** 0.4167 0.2696 0.35 0.3586 0.4833 0.1938 HSP90AA1 –0.1167 0.7756 0.6 0.0968 0.5833 0.108 0.6 0.0968** cluster 2 SERPINB1 –0.75 0.0255* 0.7833 0.0172* 0.7167 0.0369* 0.85 0.0061* ACTBL2 –0.6833 0.0503 0.65 0.0666 0.5667 0.1206 0.7333 0.0311** ACTG1 –0.6167 0.0857 0.7833 0.0172* 0.7 0.0433* 0.8333 0.0083* CORO1A –0.65 0.0666 0.6667 0.0589 0.5833 0.108 0.7333 0.0311* RSU1 –0.55 0.1328 0.7667 0.0214* 0.6833 0.0503 0.8 0.0138* MGLL –0.5833 0.108 0.8 0.0138* 0.7167 0.0369* 0.8167 0.0108* PFN1 –0.7 0.0433* 0.8167 0.0108* 0.7333 0.0311* 0.8833 0.0031** WDR1 –0.7667 0.0214* 0.9167 0.0013** 0.8667 0.0045** 0.95 0.0004*** TAGLN2 –0.6667 0.0589 0.9333 0.0007*** 0.9 0.002** 0.9 0.002** YWHAZ –0.75 0.0255* 0.7167 0.0369* 0.6833 0.0503 0.75 0.0255* ACTN1 –0.5667 0.1206 0.8167 0.0108* 0.7667 0.0214* 0.9 0.002** [113]Open in a new tab ^a Results are expressed as Pearson correlation coefficient (r) and associated p-values. * indicates p-value < 0.05; ** indicates p-value < 0.01; *** indicates p-value < 0.001; **** indicates p-value < 0.0001. Discussion In this study, we conducted a detailed analysis of MRI data, behavioral outcomes, and serum proteomics in the NHP model of ischemic stroke. Our study focused on the intricate relationship between neuroimaging findings, behavioral functions, and proteomic profiles following an ischemic stroke, aiming to integrate these diverse data streams for a holistic understanding of stroke pathophysiology. We observed that infarct volume peaked at 1-month post-MCAO and remained constant up to 3 months, a trend that correlated with changes in motor function, neurological impairment, and working memory performance. Additionally, we found significant correlations between the decline in FA and the increase in MD with both motor and cognitive deficits. The serum proteomic analysis, integrating protein identification and quantitation, revealed critical correlations with neuroimaging and behavioral findings. These findings open new avenues for a multimodal approach to understanding stroke pathology and recovery, including integrating neuroimaging, behavioral assessment, and molecular profiling. Our results revealed that infarct volume reached its maximum size 1 month after the MCAO procedure and then stabilized, showing no significant change for up to 3 months ([114]Figure [115]1 A). The structural alterations observed in the brain during the acute phase of a stroke, which include the development of infarct lesions and a decrease in manual dexterity, are likely influenced by oxidative stress and additional factors associated with reperfusion.^[116]24 These changes are reflected in our findings, where we observed a direct correlation between the extent of injury volume and the assessments of neurological function, as depicted in [117]Figure [118]4A. Moreover, we detected a significant decrease in FA and an increase in MD across several brain regions as compared to the contralateral side of the injury area ([119]Figure [120]1B). This result is consistent with previous research linking DTI metrics to injury patterns in motor fiber bundles and motor dysfunction.^[121]24,[122]25 The changes in FA and MD around the infarcted areas have significant implications for neurological functionality and the potential for recovery. Specifically, basal ganglia damage can impair motor control by disrupting the regulation of movements and coordination of learned movement patterns.^[123]26 Additionally, cortical damage can compromise cognitive processes, including memory and executive functions.^[124]27 Interestingly, FA showed signs of recovery at 3-month post-MCAO, while MD continued to increase, suggesting early stage tissue repair and axonal regrowth alongside ongoing tissue damage or inflammation ([125]Figure [126]1B). This asynchronous recovery may reflect the specificity of brain microstructure damage and recovery. Our findings corroborate previous studies on the sensitivity of FA in predicting behavioral outcomes and its potential as a clinical tool for assessing stroke prognosis.^[127]16,[128]28 One month following the MCAO procedure, the monkeys in our study exhibited significant neurological deficits ([129]Figure [130]2 ), including paralysis and reduced muscle tone in their right limbs. These deficits persisted without significant improvement over the following 3 months, illustrating the profound impact of stroke. This aligns with research linking more severe motor deficits and delayed reaction times to extensive damage in ipsilateral hemisphere.^[131]29 DTI metrics, such as FA and MD, showed significant correlations with these neurological changes ([132]Figure [133]4B). Our findings suggest that changes in DTI metrics, such as reduced fractional anisotropy or increased mean diffusivity, may be indicative of microstructural damage that has a direct impact on neuronal connectivity and function, thereby leading to the observed behavioral impairments. In addition, FA values were significantly correlated with motor function assessments, underscoring their utility in monitoring changes in motor abilities. Behaviorally, the monkeys displayed an inability to use their right upper limb for tasks like retrieving food, relying instead on their unaffected left limb ([134]Figure [135]2B). This compensatory behavior, indicative of altered motor function and spatial perception, is a well-documented response to stroke-induced impairments.^[136]18 Recent research indicates that compensatory use of the unaffected limb post-MCAO activates the contralateral motor cortex, facilitating neural reorganization and functional recovery, as observed in enhanced motor cortex activity and functional connectivity.^[137]30 Cognitively, there was a noticeable decline in working memory, as evidenced in the delayed-response task at 1-month post-MCAO ([138]Figure [139]4B). This pattern, aligning with the changes in DTI metrics, points to the impact of brain damage on cognitive functions.^[140]31 However, the observed cognitive recovery by 3 months post-MCAO highlights the brain’s resilience and capacity for recovery following such injuries. Our LC–MS/MS functional enrichment analysis identified two distinct groups, primarily involved in regulating cell morphogenesis and migration, apoptosis, and vascular functions ([141]Figure [142]3 ). Of these, 15 proteins displayed a strong correlation with DTI metrics. Notably, MMP9, THBS1, MB, and HSP90AA1 in cluster 1 were negatively correlated with FA values, while SERPINB1, ACTBL2, ACTG1, Coro1A, RUS1, MGLL, WDR1, PFN1, TAGLN2, YWHAZ, and ACTN1 in cluster 2 showed a positive correlation with MD values. These proteins also correlated positively with neurological function scores and exhibited a negative correlation with DRT correct, suggesting a connection between changes in behavioral phenotype, serum protein level, and brain pathological and functional alterations observed in MRI. MMP9, a key player in blood–brain barrier degradation, was consistently elevated in monkey serum up to 3 months post-MCAO. This sustained elevation aligns with clinical findings associating high MMP9 levels in ischemic stroke patients with increased risks of severe disability or death, underscoring its prognostic significance.^[143]32,[144]33 Our findings also indicated a increase in THBS1 post-MCAO, a protein critical in hemostasis^[145]34 and mediating cellular stress responses,^[146]35 mirroring human ischemic stroke findings where THBS1 elevation correlates with stroke severity.^[147]36 Elevated serum MB levels post-MCAO were observed, consistent with increase noted in human stroke patients^[148]37 and acute stroke monkey models.^[149]38 This sustained elevation aligns with clinical findings associating high MMP9 levels in ischemic stroke patients with increased risks of severe disability or death, underscoring its prognostic significance.^[150]39,[151]40 PFN1, an actin-binding protein involved in various physiological processes, showed elevated levels post-MCAO. Present in almost all tissues and cells,^[152]41 PFN1 is crucial for multiple physiological processes, including cell migration,^[153]42 vascular permeability,^[154]43 angiogenesis,^[155]44 and oxidative stress.^[156]45 Intriguingly, inhibiting PFN1 offered neuroprotection against ischemia/reperfusion injuries, partly by promoting M2 microglial polarization, indicating its potential as a therapeutic target in stroke management.^[157]46 YWHAZ, part of the 14–3–3 protein family, affects cellular proliferation, migration, and differentiation by interacting with phosphoserine- and serine-containing proteins.^[158]47 Clinically, elevated levels of YWHAZ poststroke have been associated with cognitive impairment,^[159]48 while experimental models demonstrate that reduced YWHAZ expression leads to neurodevelopmental and cognitive deficits.^[160]49 Our study establishes the significance of five proteins in ischemic stroke, highlighting their potential as biomarkers and therapeutic targets. Our research identified 15 proteins correlated with DTI metrics, of which 10 proteins have not been previously reported in direct association with ischemic stroke. These proteins include HSP90AA1, SERPINB1, ACTBL2, ACTG1, Coro1A, RSU1, MGLL, WDR1, TAGLN2, and ACTN1. While KEGG enrichment analyses showed their involvement in processes such as cell morphogenesis and migration, apoptosis and programmed cell death, and coagulation and blood vessels, their exact roles in the onset and recovery of ischemic stroke require further exploration. While serum proteins may not directly mirror molecular changes in brain tissue, especially in NHP models where obtaining continuous brain tissue samples is challenging, in vivo monitoring combining serum protein, behavioral analysis, and MRI remains a valuable tool for long-term disease progression tracking in animal models. In this study, we employed electrocoagulation to induce a focal cerebral ischemia model in NHPs. While this method ensures reproducibility and consistent ischemic conditions,^[161]50 it primarily models direct ischemic injury, excluding secondary reperfusion injuries common in clinical stroke scenarios. This limitation, chosen to simplify the investigation of ischemic primary effects, restricts the ability of our model to fully replicate the complex pathology of a human stroke. However, it allows for focused exploration of the immediate brain changes postischemia, leveraging advanced neuroimaging, behavioral, and proteomic analyses. Future research needs to incorporate models with reperfusion injury for a comprehensive understanding of stroke pathology and recovery mechanisms. In summary, our study of an NHP model of ischemic stroke provides key insights into the complex dynamics of stroke, linking neuroimaging data with behavioral and proteomic changes. We highlighted the significance of certain serum proteins, including MMP9, THBS1, MB, PFN1, and YWHAZ, as potential biomarkers and therapeutic targets. Our integrative approach underscores the importance of combining neuroimaging, behavioral studies, and proteomics for a comprehensive understanding of stroke, offering promising avenues for enhancing stroke research and treatment strategies. Materials and Methods Animal Preparation and Surgery Three male cynomolgus macaques (crab-eating macaque 3050703, 13050217, and 13051415), aged 8.67 ± 0.57 years and weighing 8.38 ± 1.99 kg, were included in this study. Housed individually under a 12-h light/dark cycle, the monkeys were maintained in an environment controlled at 24 ± 2 °C with 40 ± 20% humidity. They had free access to water and were fed twice daily. The MCAO model adapted from established protocols^[162]50 was accomplished in the three monkeys using bipolar electrocoagulation. Comprehensive evaluations, including MRI data collection, behavior assessments, and serum proteomic analyses, were systematically conducted. These evaluations were performed at three time points: before the induction of MCAO to establish a baseline and then at 1 month and 3 months following the MCAO procedure. All procedures adhered to the “Care and Use of Laboratory Animals” guidelines and were approved by the Institutional Animal Care and Use Committee of Guangdong Laboratory Animals Monitoring Institute (IACUC no. 2018014, AAALAC accredited). For MCAO surgery, anesthesia was initiated with ketamine (10 mg/kg), xylazine (2 mg/kg), and atropine (2 mg/kg) intramuscularly and maintained with 1–1.5% isoflurane vaporized in 100% oxygen. Vital signs, including heart rate, blood pressure, respiration, and oxygen saturation, were continuously monitored. The surgery involved aseptic removal of the left frontotemporal bone, exposure of the Sylvian fissure, and occlusion of the left MCA distal to the M1 branch. Postsurgery, the site was washed with saline, treated with bone wax, and sutured. Then, the monkeys were housed in their cages with regulated body temperature for recovery. The first 3 days post-MCAO required three-daily hand-feeding due to unilateral impairments. Antibiotics (penicillin, 0.4 million units, intramuscularly) and mannitol (20%, intravenously) were administered twice daily for infection prevention and intracranial pressure reduction. A veterinarian monitored daily until independent self-care was reestablished. MRI Image Acquisition and Data Processing Using a 3.0 T MRI system (MAGNETOM Trio, Siemens Healthineers, Erlangen, Germany) equipped with an 8-channel head coil, we acquired MRI data from each anesthetized animal (maintained with 2.5% isoflurane). Animals were placed in an MR-compatible stereotactic apparatus for imaging. Structural images were obtained using a T2-weighted turbo spin echo sequence (TR = 3000 ms, TE = 396 ms, voxel size: 0.6 × 0.6 × 0.6 mm, and BW = 651 Hz). Accordingly, diffusion tensor imaging (DTI) was performed to provide microstructural images using a multiband echo-planar sequence (TR = 5700 ms, TE = 99 ms, voxel size: 1.3 mm × 1.3 mm × 1.3 mm, and BW = 1502 Hz). The infarct volume resulting from MCAO was manually outlined on T2-weighted images (T2WI) at 1-month and 3-month post-MCAO using ITK-SNAP software (version 3.8, [163]http://www.itksnap.org). This software automatically calculated the infarct volume by summing the volumes of identified lesions. Lesion size was determined by the ratio of the lesion volume to total brain volume. We defined regions of interest (ROIs) on T2WI using the Paxinos Macaque Atlas^[164]23 (available at [165]https://scalablebrainatlas.incf.org/macaque/CBCetal15), focusing on infarcted areas. The average DTI values for the entire ROI were calculated. ROIs showing no significant differences were excluded, sharpening our focus on areas impacted by ischemia. This processing refined our analysis, enhancing the precision and applicability of our findings to understanding stroke pathophysiology. DTI data were processed with FSL (version 6.0, FMRIB Software Library, [166]http://www.fmrib.ox.ac.uk/fsl) and DSI Studio (version 12, Yeh FC. [167]https://dsi-studio.labsolver.org). DTI metrics, including fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radical diffusivity (RD), were measured to evaluate microstructural properties and diffusion characteristics in the brain.^[168]51 Behavioral Assessment This study conducted a comprehensive behavioral assessment in nonhuman primates including neurologic deficit evaluation, motor function analysis (hill and valley staircase task, HVST), working memory assessment (delayed-response task, DRT), and cognitive function evaluation (object-retrieval detour task, ORDT). These assessments were conducted before MCAO induction and at 1-month and 3-month post-MCAO. Neurologic deficits were quantified using a standardized neurological deficit score,^[169]52 including four categories totaling 100 points: consciousness (28 points), sensory system (22 points), motor system (32 points), and skeletal muscle coordination (18 points). Two blinded, experienced observers performed these assessments, with a higher score indicating more server neurological damage. We assessed motor function and spatial perception using HVST.^[170]53 Monkeys retrieved food rewards from staircases outside their cages, reaching through vertical slots in a front-panel plexiglass cage. Scores were assigned based on the distance from the slot (1 for nearest, 5 for furthest), with a maximum possible score of 15 per side. In the hill task, there are two staircases positioned such that they converge toward the center at the top. This arrangement resembles a hill. Conversely, in the valley task, the staircases are arranged to diverge from a central point like a valley. Monkeys were given 3 min to retrieve all the food items (apple pieces) from the staircases. The order of testing was randomized to prevent bias. The maximum possible score of each staircase was 45. Working memory was evaluated using DRT with the Wisconsin General Test Apparatus.^[171]54 Monkeys were trained on a two-well delayed response task before surgery. During the actual test, a piece of food was randomly and visibly placed in one of the two wells and then both wells were covered with lids. An opaque clapboard was then placed between the monkey and wells, obstructing the view. After a predetermined delay, the clapboard was removed, and the monkey was allowed to select one of the two lids to lift in search of the food reward. The final set of delays includes 0, 4, 8, 12, and 16 s. The average correct response rate over five rounds indicated working memory proficiency. Cognitive function was assessed using ORDT.^[172]55 Monkeys attempted to retrieve food from a transparent plexiglass box with varying open-side orientations. The test comprised 17 trials, with both easy (9 trials) and difficult (eight trails) food retrievals. Performance was quantified by the success rate (first attempt success) and number of barrier hits (attempts to reach the reward through the transparent wall). Each monkey underwent five rounds with an average success rate and barrier hits assessed. Proteomics Blood samples were obtained from three animals before MCAO surgery and at 1-month and 3-month post-MCAO. Serum was extracted from these samples and stored at −80 °C. Proteins in each serum sample were denatured, reduced, alkylated, and digested with sequencing-grade modified trypsin (protein-to-enzyme ratio of 50:1) at 37 °C overnight. Then, peptides were recovered through centrifugation (12,000g for 10 min) and desalted using a Strata X SPE column. Peptides, dissolved in solvent A, were loaded onto a homemade reversed-phase analytical column (25 cm length, 100 μm i.d.). A gradient consisting of solvent A (0.1% formic acid and 2% acetonitrile in water) and solvent B (0.1% formic acid and 90% acetonitrile in water) was employed for peptide separation on an ASY-nLC 1200 UPLC system (ThermoFisher Scientific). The separation gradient was set as follows: 0–68 min, 4%–20% B; 68–82 min, 20%–32% B; 82–86 min, 32%–80% B; and 86–90 min, 80% B, maintaining a constant flow rate of 500 nL/min. The peptides were then analyzed using an Orbitrap Exploris 480 instrument with a nanoelectrospray ion source (electrospray voltage: 2300 V). FAIMS compensate voltage (CV) was set at −70 V, −45 V. The Oorbitrap detector was configured for optimal analysis with a full mass spectrometry (MS) scan resolution of 60,000 (scan range: 400–1200 m/z) and MS/MS scan fixed at a first mass of 110 m/z, resolution 30,000. TurboTMT was disabled. A maximum of 15 abundant precursors were selected for MS/MS analyses with a 30 s dynamic exclusion. HCD fragmentation was performed at a normalized collision energy (NCE) of 27%, with an automatic gain control (AGC) target of 75%, an intensity threshold of 10,000 ions/s, and a maximum injection time of 100 ms. The resulting MS/MS data were processed using the Proteome Discoverer search engine (v.2.4), searching against the Macaca_fascicularis_9541_PR_20220314.fasta database (50208 entries). Search parameters included Trypsin (Full) as the cleavage enzyme, allowing for up to 2 missed cleavages, a minimum peptide length of 6, and a maximum of a modifications per peptide. Mass error setting was 10 ppm for precursor iron and 0.02 Da for fragment ions. Fixed modification was carbamidomethyl on Cys, while variable modifications included oxidation on Met, acetylation on protein N-terminal, met-loss on Met, and met-loss + acetyl on Met. False discovery rate (FDR) was adjusted to less than 1% for protein, peptide, and PSM levels to ensure high data confidence. For gene ontology and pathway analysis, the org.Hs.eg.db R package from Bioconductor was used to correlate UniProtKB accession and gene name with ENTREZ Gene IDs. Functional classification and enrichment analyses of gene ontology (GO) terms among differentially expressed proteins were performed by using ENTREZ Gene IDs via the clusterProfiler R package. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis was also conducted using the clusterProfiler R package, supported by a bioconductor package for pathway-based data integration and visualization. Statistical Analysis Statistical analyses were conducted using GraphPad Prism (ver. 9, GraphPad Inc., San Diego, CA, USA). Data were represented as the mean ± standard error of the mean (SEM). A two-sample t test was employed to compare ipsilateral and contralateral changes at baseline (before MCAO), and at 1-month and 3-month post-MCAO time points. Difference across the three time points (baseline, 1-month, and 3-month post-MCAO) was analyzed using two-way ANOVA and was carried out with three time points, followed by post hoc analysis using Tukey’s test to identify specific group differences. Based on the assumption that ischemic brain events trigger systemic biological and behavioral responses, we analyze the correlation between DTI metrics, serum protein levels, and behavioral assessment to determine biomarkers for stroke using Spearman correlation coefficients. To integrate multiple biomarkers, including DTI metrics, behavioral scores, and serum proteomics, a simple linear regression analysis was employed. This approach was intended to provide a more comprehensive understanding of the interplay between different types of data. Acknowledgments