Stroke Proteomics: From Discovery to Diagnostic and Therapeutic Applications

Abstract

Stroke remains a leading cause of death and disability, with limited therapeutic options and suboptimal tools for diagnosis and prognosis. High throughput technologies such as proteomics generate large volumes of experimental data at once, thus providing an advanced opportunity to improve the status quo by facilitating identification of novel therapeutic targets and molecular biomarkers. While proteomics studies in animals are largely designed to decipher molecular pathways and targets altered in brain tissue after stroke, studies in human patients primarily focus on biomarker discovery in biofluids and, more recently, in thrombi and extracellular vesicles. Here, we offer a comprehensive review of stroke proteomics studies conducted in both animal and human specimen, and present our view on limitations, challenges, and future perspectives in the field. In addition, as a unique resource for the scientific community, we provide extensive lists of all proteins identified in proteomic studies as altered by stroke, and perform post-analysis of animal data to reveal stroke-related cellular processes and pathways.

1. Introduction

Stroke is a highly prevalent and often devastating disease that primarily affects the elderly¹. It is a leading cause of death and long-term disability, presenting a major socioeconomic burden. In the past several decades, significant progress has been made in stroke prevention and acute stroke care, while only limited advances were achieved in identifying biomarkers to assist diagnosis/prognosis and developing therapies to counteract the harmful effects of stroke².

There are two main stroke subtypes, ischemic and hemorrhagic, which arise from cerebral vessel blockage and bleeding into the brain parenchyma (intracerebral hemorrhage, ICH) or subarachnoid space (subarachnoid hemorrhage, SAH), respectively. Patients suffering from either stroke type often present with similar clinical features, making early accurate diagnosis challenging. Current stroke diagnosis almost exclusively relies on a patient’s neurologic presentation and neuroimaging, which is time-consuming, and requires highly experienced neurologists and advanced imaging facilities that are often not accessible. Identifying a rapid, reliable, and inexpensive stroke biomarker may greatly help to quickly diagnose stroke and stroke subtypes in patients, especially in resource-limited medical facilities, and ensure that they receive timely and appropriate medical care. In addition, therapeutic options following stroke are extremely limited. The treatment of ischemic stroke (IS) relies primarily on acute thrombolysis with tissue plasminogen activator (t-PA) and/or mechanical thrombectomy to restore blood flow, and there is no specific medication for hemorrhagic stroke (HS). Finding new strategies to increase post-stroke brain cell survival is essential, and will only succeed if we identify better targets through a deeper mechanistic understanding of stroke pathophysiology.

The dawn of the proteomics age has allowed researchers to obtain a comprehensive picture of the protein landscape altered in brain tissue, bodily fluids, and other biological materials following stroke. If utilized to its full potential, this information can greatly increase the chances of discovering stroke biomarkers and therapeutic targets. Proteomics is the large-scale analysis of the entire set of proteins (termed proteome) in a biologic system at a given time. Proteins carry out most essential cellular functions. Therefore, the proteome’s composition bears critical information related to an organism’s state. The proteome is highly dynamic and adapts in response to changes in the environment through transcriptional, translational, post-translational, and degradational processes. Proteomics is advantageous over other high throughput omics techniques as it assesses the end-product arising from a combination of adaptations on every level. This is particularly relevant to stroke, after which translation is stalled³ and altered mRNA levels, detected by transcriptomics, may not correlate with similar changes in protein levels. Additionally, post-translational protein modifications (PTMs), which cannot be investigated by any other omics method, may be important modulators of stroke outcome.

To date, stroke proteomics has used different sample sources by global or targeted mass spectrometry (MS) approaches as well as antibody- and aptamer-based arrays (Figure 1). Here, we summarize proteomics data from animal and human stroke studies, discuss their potential value for stroke diagnosis/prognosis and therapy, with focus on biomarker and target discovery, and provide thoughts about future research in this field.

Figure 1.

Schematic diagram summarizing sample sources, proteomics methods and applications of stroke proteomics studies executed in animal models and patients. AP, affinity purification; CSF, cerebrospinal fluid; ECF, extracellular fluid; EVs, extracellular vesicles; ICH, intracerebral hemorrhage; MS, mass spectrometry; SAH, subarachnoid hemorrhage. Image was partly created with BioRender.com.

2. Proteomics from brain tissue: in search of disease mechanisms and therapeutic targets

2.1. Animal studies

Global proteomic changes in the ischemic core and penumbra during acute, subacute, and chronic stroke phases were assessed in rodent, porcine, and nonhuman primate models of IS and HS to identify proteins involved in brain cell death, survival, and recovery. A list of all animal studies containing details about type of animal and stroke model used, sampling sources and timepoints, as well as lysis and proteomics methods used for protein identification can be found in Table S1. Over 20 studies identified >7,000 known and new differentially expressed proteins (DEPs) in different brain areas at various timepoints after stroke (DEPs listed in Table S2). Here, we provide a review of the major targets and associated molecular pathways identified across all studies. To determine pathways consistently altered throughout all post-stroke phases in each stroke type, we pooled the available data, and performed pathway and gene ontology (GO) enrichment analyses with DEPs that appeared in at least two proteomics studies (Table S3, sheet 1). Identified DEPs suggest changes in platelet regulation and complement/coagulation cascades in both stroke types, while protein phosphorylation and neuron-specific pathways were only altered by ischemia (Figure 2A). DEPs connected with platelets and coagulation, as well as phosphorylation, were upregulated, indicating heightened responses, while those associated with synaptic function were reduced (Figure 2B). Enriched GO terms among decreased DEPs after ischemia reiterated deficient neuronal function, while an increase in enzymatic activity, metabolism, and proteolysis was observed (Figure 2C). DEPs after HS were exclusively upregulated and mainly associated with extracellular material, regulation of enzymatic activity, and coagulation (Figure 2C). Below, we describe selected DEPs and protein classes that were consistently changed in different stroke phases across multiple proteomics studies. An omitted protein, however, should not be considered insignificant.

Figure 2.

Heatmaps showing (A) (B) KEGG and Reactome pathways and (C) GO terms significantly enriched among DEPs in animal brain tissue after IS and HS. Datasets were derived from the following studies: Ischemia^{4–8,11–16,22–24,30,32,33,36–38,55,60,71}, ICH^6,39–41,50. Only DEPs that were identified in at least 2 proteomics studies were considered for analysis (Table S3, n=293 for ischemia and n=47 for ICH). Functional enrichment analysis was performed using g:Profiler¹⁶⁶. Benjamini-Hochberg FDR-corrected P values were plotted on a Log₁₀ scale. P values were considered significant for (A) (B) P<10⁻⁵ and (C) ischemia: GO-MF P<10⁻¹², GO-BP P<10⁻¹³, GO-CC P<10⁻²⁰; ICH: GO-MF P<10⁻⁷, GO-BP P<10⁻⁵, GO-CC P<10⁻⁹. FDR, false discovery rate.

2.1.1. IS - Acute phase

DEPs identified via proteomics using rodent tissue obtained after IS and 1–12h reperfusion confirm altered transcription, translation, protein phosphorylation, organelle function, and neurotransmission in this early phase after ischemia. For example, middle cerebral artery occlusion (MCAO; a commonly used IS model in animals) and 1 hour (h) reperfusion in mice significantly changed the abundance of histones, RNA-binding proteins (FXR2, FUS, SFPQ), ribosomal subunits, kinases (PKC, ITPKA, adenylate kinase-5), calcium-binding proteins (annexin-2, annexin-6), and ion channels (SLC6A11, VDAC1) in the ischemic penumbra⁴. Another proteomics study reported a similar trend toward upregulation of kinases (PKA, PKC), and RNA-binding and transcription factors (STAT6, PRP16, FUS), whereas neuronal proteins (PSD95, MAP2), and proteins associated with cellular adhesion and vesicular transport (clathrin heavy chain, Ras-related protein, paxillin) exhibited downregulation after MCAO and 0, 3, and 12h reperfusion in mice⁵. A third study performed in rats confirmed differential regulation of kinases (CAMK2A, phosphoglycerate kinase-1, PKC) and synaptic regulators (synaptophysin, synapsin-2, NSF, MAP6), as well as metabolic enzymes, such as adenosine kinase-1, adenosyl homocysteinase, alcohol dehydrogenase-1, and glutamate dehydrogenase-1, in the ischemic hemisphere after MCAO/3h reperfusion⁶. After MCAO/0.5h reperfusion, similar DEPs as in the penumbra were detected in the mouse ischemic core, with particular emphasis on mitochondrial respiration (NDUFA1, NDUFS5, UQCRQ) and vesicular transport (VAMP2, ATP6V1F)⁷. Histones, however, were not altered in the core⁷. Two DEPs from this study, GADD45G and catenin δ−2, were increased in circulating blood of human stroke patients, where they might prove promising as stroke prognostic and diagnostic biomarkers⁷.

Proteomics executed after 24–48h of permanent or transient ischemia in rodents identified DEPs that reflect post-stroke alterations in protein homeostasis, energy metabolism, and synaptic function/plasticity. 24h of MCAO significantly upregulated DRP2, HSC70-ps1, and SPNA-2 fragment in the ischemic penumbra in rats⁸. DRP2, a protein governing axonal growth, and synaptic plasticity and maturation^9,10, was upregulated after 24h of MCAO in several other proteomics studies in rats^11–15, presenting itself as a reliable marker for the ischemic brain. Interestingly, DRP2 upregulation appears to be restricted to MCAO without reperfusion, suggesting a normalizing effect of blood flow restoration. In contrast, proteomics found elevated SPNA-2 fragment levels in rodents after MCAO with and without reperfusion^8,16. Fragmentation of SPNA-2, a protein required for maintaining neuronal membrane integrity, indicates ischemic stress and foreshadows neuronal cell death¹⁷. The products of SPNA-2 breakdown were also detected in the cerebrospinal fluid (CSF) of MCAO-treated rats^18,19 and in patients with subarachnoid hemorrhage²⁰, demonstrating the potential of SPNA-2 as a biomarker for IS and HS.

Heat shock proteins (HSPs) are well-known to be impacted by ischemia in the acute phase²¹, a finding that proteomics studies confirm. Both HSC70 and HSP60 were elevated after 24h of MCAO in the rat cortex^8,11, while HSP90 and HSPA5 were decreased after MCAO and 24h reperfusion in the rat peri-infarct area²². Additionally, HSPs, specifically the inducible form of HSP70²³, as well as HSP10 and HSP70¹⁶, were increased in rats after 48h MCAO with and without reperfusion. Proteomics also found that DRP2 and HSC70, among others, were increased in pituitary glands in rats after 24h of MCAO²⁴, suggesting a common response to injury in cortical and neuroendocrine systems.

Thioredoxin, peroxiredoxin-2 and UCH-L1 are DEPs commonly detected by proteomics as downregulated in rodents in the acute stroke phase^11,16. Lower levels of thioredoxin and peroxiredoxin-2, two key molecules in combating ischemic oxidative stress^25,26, and UCH-L1, a deubiquitinase implicated in neuroprotection²⁷, may actively exacerbate ischemic cell death. However, on the other hand, reduction in these and other identified proteins could also be only a consequence of already progressed tissue damage and death. While intracellular UCH-L1 decreased after IS and traumatic brain injury (TBI), elevated levels were found in the CSF of rats, warranting further investigation as a biomarker for certain types of brain injury^19,28. Indeed, a serum biomarker test combining measurements of UCH-L1 and GFAP within 12h of head injury was recently approved by the FDA for diagnosing mild TBI²⁹. Notably, levels of some DEPs, such as SPNA-2 fragment, HSP10, and HSP70, were altered regardless of occlusion time (0.5h vs 2h MCAO), while reduction in UCH-L1 and HSP90 was exclusively seen with longer occlusions¹⁶, potentially representing markers for ischemia severity. Other notable DEPs identified by proteomics in the acute phase in the ischemic penumbra are tropomodulin-2, calmodulin-1, superoxide dismutase-1, synapsin-1, MDH2, stathmin, and apolipoprotein-A1^8,11,22.

Proteomic analysis of the ventricular zone, which primarily contains migrating neurons, radial glial cells, and progenitor cells, revealed significantly increased levels of antiproteases and cytokine/growth factor-binding proteins, including α−2-macroglobulin, albumin, apolipoprotein-A1, SERPINA3K, and transferrin, in mice after 4h of MCAO³⁰. Interestingly, in the penumbra these protein classes tend to be altered in the subacute and chronic phases of ischemia, which is described in the next section.

2.1.2. IS - Subacute and chronic phases

Proteomics of the subacute and chronic phases of ischemic injury identified DEPs that potentially modify post-stroke synaptic function, lysosomal activity, and the inflammatory response, which plays a critical role in late-stage stroke pathophysiology³¹. After MCAO followed by 7 or 14d reperfusion, cathepsin and LAMP2 levels were increased in the ischemic penumbra of rodents, as were plasma, and complement and immune mediators such as complement C3, C-reactive protein, albumin, fibrinogens, annexin-2, transferrin and α−1-antitrypsin³². This suggests enhanced lysosomal activity and an activated inflammatory response. Proteomics and subsequent immunohistochemistry of the rat cortex after MCAO with 3 or 14d reperfusion showed increased expression of annexin-3, specifically in microglia³³, revealing a new marker that has since been used to detect activated migrating microglia in the ischemic penumbra³⁴. Large-scale analysis of the cortical proteome following permanent distal hypoxia-MCAO in mice identified many secreted DEPs such as apolipoprotein-E, α−2-macroglobulin, albumin, SERPINA3K and apolipoprotein-A1, at 1, 3, 7, 14, and 28d³⁵, again indicating affected immune responses, coagulation, and plasmin signaling. A similar DEP profile was discovered in the mouse cortex after MCAO and 0, 6, 12, or 24h reperfusion³⁶. Identified proteins included serpins, fibrinogens, albumin, apolipoprotein-A1, C3, S100A9, transthyretin, and transferrin, which were upregulated as early as 6h after onset of reperfusion. After global ischemia in mice, induced by bilateral common carotid artery occlusion (BCCAO), and 72h reperfusion, cortical, hippocampal, and striatal proteins involved in inflammation and oxidative stress (apolipoprotein-E, SERPINE2, clusterin, cystatin-C) were increased, while proteins regulating synaptic function (α-synuclein, synaptopodin, CAMK2A) were decreased³⁷. Clusterin and cystatin-C were also elevated in plasma of IS patients within 24h of stroke onset³⁷, warranting further investigation as blood biomarkers for acute IS. Substantiating rodent studies, proteomics revealed similar DEP classes in the penumbra of nonhuman primates after MCAO and 28d reperfusion³⁸.

2.1.3. ICH – primary and secondary injury phases

ICH led to a significant upregulation of plasma proteins in the parenchyma of rodent brains. DEPs such as albumin, α−2-macroglobulin, apolipoprotein-A1, C3, fibrinogens, hemoglobin, haptoglobin, hemopexin, ITIH3, plasminogen, S100A9, and serpins were elevated in the primary injury phase after 3h⁶, and in the secondary injury phase at 24h³⁹, 48h⁴⁰, and 72h⁴¹ in rats and mice, attesting to an early and prolonged reaction to ICH. Parenchymal apolipoprotein-A1, albumin, α−2-macroglobulin, and serpins may contribute to hematoma clearance through heightened hemostasis and platelet degranulation, while elevated C3 and hemoglobin aggravate brain injury through excessive recruitment of immune cells⁴² and production of cytotoxic heme and iron⁴³, respectively. Induction of hyperglycemia in rats led to augmented neuronal apoptosis 12h and 24h after ICH, which was accompanied by a reduction in albumin and other DEPs⁴⁴. Low serum albumin levels were associated with poor functional outcome in patients with SAH, ICH, and IS^45–47 by increasing the risk of for infections and directly aggravating brain injury^48,49. The role of other identified plasma DEPs in ICH injury development remains unclear, and requires further assessment.

Contrary to plasma mediators, levels of intracellular proteins were reduced after ICH. Metabolic enzymes such as alcohol dehydrogenase-1, ATP-citrate synthase, ALDOC, glutamate dehydrogenase-1, lactoylglutathione lyase-1, L-lactate dehydrogenase-A, PFKM, and transketolase exhibited decreased levels at 3h after ICH, suggesting impaired metabolic performance early after the insult. The dysregulated metabolism was maintained after 48h and 72h^40,41. Most enzymes were still decreased (e.g., ALDOC, L-lactate dehydrogenase-A, GOT1, ECHS1, NDUFA9, UQCRB, ATP5C1), while some were increased (e.g., ATP1A1, carbonic anhydrase-1, ITPKA, PFKM, copine-7). This demonstrates a bidirectional regulation of energy metabolism at this stage after ICH. Other intracellular protein classes with altered abundance in rodent and pig brain tissue after ICH included septins, neurofilaments, solute carriers, serine-threonine protein phosphatases, kinases, proteasomal subunits, HSPs, and synaptic proteins, among others^6,39–41,50.

2.1.4. IS – cytoprotection

Proteomic analyses of tissue from animals protected from ischemic insult through genetic advantage, preconditioning or treatment with a cytoprotective agent aim to identify proteins that aid in protection from injury. We performed GO enrichment analysis of DEPs changed in vulnerable and protected tissue after focal IS in at least two studies (listed in Table S3, sheet 2). Protection led to a distinct upregulation of DEPs associated with preservation of neuronal and synaptic integrity, while largely normalizing changes observed in the vulnerable condition (Figure 3). The potential role of select DEPs in cytoprotection is described below.

Figure 3.

Heatmap comparing GO terms significantly enriched among post-stroke DEPs in animal brain tissue vulnerable to and protected from focal IS (MCAO). Datasets used for analysis were from the following studies: Vulnerable^{4–8,11–16,22–24,30,32,36,38,55,60,71}, protected^{55,60,64,67,70,71,73}. Enrichment analysis and data plotting were performed as in Figure 2. P values were considered significant for GO-MF P<10⁻⁸, GO-BP P<10⁻¹⁰, GO-CC P<10⁻⁸; n=98 for vulnerable UP, n=135 for vulnerable DOWN; n=8 for protected UP, n=7 for protected DOWN). FDR, false discovery rate; MCAO, middle cerebral artery occlusion.

2.1.4.1. Estrogen

Females are protected from IS during their reproductive years. Proteomics from ovariectomized female rats pretreated with vehicle or estradiol found that estradiol prevented the increase in HSP60 levels after 24h of MCAO observed in vehicle controls⁵¹. Meanwhile, PEA15 and PPP2A were elevated in estradiol-treated vs non-treated animals. In PEA15 knockout mice susceptibility of astrocytes to TNF-induced apoptosis was increased⁵²; hence, reduced PEA15 could favor stroke-mediated apoptosis, while decreased PPP2A may contribute to ischemic injury by increasing pTau⁵³. Interestingly, decreased HSP60 and increased PEA15 and PPP2A were already present in estradiol-treated animals before MCAO⁵¹, indicating that they could underlie the mechanism that protects females against stroke.

Additionally, estradiol protects against ischemia by increasing neuronal Stat3 activity⁵⁴. To identify potential mediators of this effect, proteomics analyzed male and female neuronal Stat3 wild type and knockout mice after MCAO and 4d reperfusion. Stat3- and sex-specific DEPs consisted of metabolic (ACO2, superoxide dismutase-2, γ-enolase), synaptic (α-synuclein, synapsin-2), structural (14–3-3 protein β/α, profilin-2, DRP2), and transcriptional (Pur-α, BASP1) regulators⁵⁵. Their involvement in the Stat3-dependent neuroprotection in females, however, has not yet been determined. One caveat of this study is that the proteome was assessed 4d after MCAO, thus differing extents of cell death in experimental groups could confound results.

2.1.4.2. Ischemic preconditioning

Exposure to a short episode of ischemia induces protection against a subsequent severe ischemic insult. Proteomic changes in response to ischemic preconditioning could explain the molecular mechanisms underlying the induced tolerance. At 24h after 10min sublethal MCAO, HSP70 and HSP27, whose overexpression results in less ischemic brain damage^56,57, were upregulated in the rat cortex⁵⁸. Other proteins that increased by preconditioning were guanylyl cyclase, PTAFR, β-actin, and muskulin-1⁵⁸. Guanylyl cyclase knockout mice exhibited increased infarct volume after MCAO⁵⁹, hence upregulated guanylyl cyclase is expected to attenuate stroke injury. Noteworthy, the above study was executed in spontaneously hypertensive rats, and the extent to which hypertension influenced the proteome is unclear.

Another proteomics study found a distinctive increase in transcriptional repressors, particularly in polycomb group (PcG) proteins, in tolerant vs vulnerable mice after MCAO and 24h reperfusion⁶⁰. This is one of the rare proteomics studies in which an effect of the identified proteins on outcome was confirmed. PcG protein knockdown precluded induction of ischemic tolerance, whereas exogenous overexpression conferred tolerance without the preconditioning stimulus. PcG proteins conferred protection by decreasing abundance of potassium channels, whose activity is detrimental to stroke outcome⁶⁰. In addition to PcG proteins, ischemic preconditioning also impacted levels of certain ATPase subunits, histones, kinases, phosphatases, and Ras-related proteins, all with currently unknown importance in preconditioning.

2.1.4.3. Pre-injury treatment with cytoprotective agents

Pretreatment with the non-erythropoietic derivative asialoEPO 4h before insult led to robust protection against hypoxia-ischemia in neonatal rats⁶¹. Proteomics showed pre-insult induction of cortical p-SNAP25⁶¹, which may contribute to post-ischemic neuroprotection by priming SNARE complex formation and vesicle exocytosis^62,63.

Fenofibrate and atorvastatin are hypolipidemic, neuroprotective drugs with pleotropic and partly overlapping functions involving PPARα induction. Administering these drugs 14d before MCAO in rats affected pre- and post-ischemic cortical levels of α-synuclein, 14–3-3 protein ζ/δ, and protein disulfide-isomerase⁶⁴, which could confer protection by preserving synaptic and endoplasmic reticulum (ER) function.

2.1.4.4. Post-injury treatment with cytoprotective agents

Nicotinamide, a B₃ vitamin and precursor to NAD⁺, protects against ischemic injury by acting as PARP inhibitor and scavenging of free radicals. Like in females⁵¹, proteomics found that, administration of nicotinamide 2h after occlusion prevented the increase in HSP60 and decrease in PEA15 and PPP2A observed in vehicle-treated rats after 24h of MCAO¹⁴. The same alterations of HSP60, PEA15 and PPP2A levels were also reported after ferulic acid treatment in rats after 24h of MCAO¹³. In addition to these proteins, nicotinamide increased cortical levels of NSE¹⁴; however, the meaning of this increase is unclear due controversial findings regarding post-stroke NSE tissue levels^16,32. Ferulic acid on the other hand restored cortical levels of MAP2K1, GAPDH, IDH2 and adenosyl homocysteinase¹³, which could provide protection by maintaining energy metabolism and preserving adenosine levels.

Post-injury application of pituitary adenylate cyclase activating polypeptide (PACAP)38, a neuroprotective peptide resembling a hypothalamic hormone, promotes an increase in DRP2 levels at 6h rather than at 24h of MCAO in mice^{8,11–15,24,65}, which could accelerate neurorepair^9,10.

Post-treatment with the anti-inflammatory and antioxidant compound resveratrol restored cortical UCH-L1 levels in the rat after 24h of MCAO¹², which could directly contribute to protection via promoting axonal repair⁶⁶. Additionally, it restored peroxiredoxin-5 and IDH levels¹², potentially relieving oxidative and mitochondrial stress.

Proteomics assessment after treatment with simvastatin 15min after occlusion revealed elevated GNAO1 and MAPRE, and lower HSP60, HSP75, and fumarate hydratase after 48h of MCAO⁶⁷. HSP75 was also reduced in plasma of simvastatin-treated patients on days 3 and 5 after stroke⁶⁷. While it may show promise as a biomarker, reduced HSP75 is not likely to have a beneficial effect on injury development, as HSP75 overexpression proved protective against ischemic injury⁶⁸. Rather, decreased HSP75 may reflect reduced injury resulting from a mechanism involving another affected protein such as GNAO1. Since GNAO1 plays key roles in transducing GPCR signals to regulate neuronal excitability⁶⁹, restoring GNAO1 function to improve stroke outcome should be further explored.

Proteomics revealed that post-treatment with rosiglitazone, a PPAR-γ-dependent neuroprotectant, in rats restored 14–3-3 protein ε levels after MCAO and 24h reperfusion⁷⁰. This is one of the few studies with follow-up showing that the identified protein indeed underlies protection. While 14–3-3 protein ε siRNA knockdown abrogated rosiglitazone-dependent protection, its overexpression was sufficient to alleviate infarct volume without treatment⁷⁰.

Administration of the anticoagulant t-PA 4h after occlusion in rats altered cortical levels of many proteins involved in excitatory neurotransmission and signaling after 24h reperfusion⁷¹. Examples include glutamate decarboxylase-2, EAAT1, neurofilaments, MAPT, CAMK2D, Ras-related protein-10, and α-internexin. These changes may underlie some of the neurotoxic effects of t-PA seen after ischemia⁷².

Treatment with the reactive oxygen species (ROS) scavenger tetramethylpyrazine nitrone (TBN) 3h after MCAO in nonhuman primates conferred neuroprotection, which on the proteomic level, was associated with reduced mitochondrial and redox-sensitive proteins (e.g., COX5B, superoxide dismutase-1, and peroxiredoxin-2) and elevated cytoskeletal and tight junction proteins (claudin-11, MBP, and neurofilaments) at 28d reperfusion⁷³.

2.1.4.5. Hypothermia

Hypothermia is one of the best-studied protective therapies for brain ischemia. Proteomic changes in response to hypothermia in rats included normalization of cortical DRP2 levels after 24h of MCAO¹⁵, with unknown consequences. Hypothermia also increased availability of BAIAP2L1 and profilin-2, perhaps to stabilize the cytoskeleton, and MAP2K1 and PPIA, to support metabolic and anti-apoptotic functions¹⁵. Proteomic changes in this study were observed only in the ischemic penumbra, not in the infarct core, signifying that protection of the salvageable area of the brain may involve one or more of the identified proteins.

2.1.5. IS and preconditioning - changes in protein interactions, modifications, and aggregation

Protein function is regulated not only by changes in its levels but, also by its interaction profile with other proteins, its subcellular localization, and its conjugation with posttranslational modifiers. Identifying such alterations in response to a preconditioning stimulus or stroke may increase our mechanistic understanding of stroke pathophysiology and aid in finding therapeutic targets. In this section, we describe studies that used affinity purification-coupled MS (AP-MS) to analyze preconditioning- and stroke-dependent changes in protein interaction complexes of PKC, PPP1 and 4E-BP2, PTMs such as SUMOylation and ubiquitination, and aggregation propensity of proteins.

2.1.5.1. Protein interactions

PKC and PPP1 activities are essential for hypoxic and ischemic preconditioning^74–77. Several proteomics studies were executed to decipher the interactome around these important proteins in the presence and absence of preconditioning. PKCβ and ε bound to 49 and 39 proteins, respectively, in cortices of control and hypoxic preconditioned mice, revealing an extensive interaction network^75,76. Notably, β and ε PKC showed a distinct, largely non-overlapping interaction profile, reflecting different roles. In response to preconditioning, ten proteins were differentially bound to PKCβ. Their identity predicted a role for PKCβ in preconditioning by altering energy metabolism, oxidative stress, protein folding and degradation, and axon guidance⁷⁵. DRP2 was confirmed to be a novel PKC phosphorylation target⁷⁵. Preconditioning led to increased DRP2 binding to PKCβ localized at the membrane, thereby preventing DRP2 dephosphorylation and cleavage after ischemia, which could enhance axonal resistance to neurotoxicity⁷⁸. Eight proteins exhibited altered binding to PKCε after preconditioning (i.e., triosephosphate isomerase-1, HSC71, GAPDH, HSP60, GRP78, 14–3-3 protein γ, Pur-α, and ACO1)⁷⁶. Potential effects of any of these changes on PKCε-dependent preconditioning have not yet been determined. According to current knowledge, PKCε-dependent preconditioning is achieved through positive regulation of mitochondrial respiration⁷⁹.

AP-MS proteomics identified PPP1α- and γ-interacting proteins that significantly changed association in response to ischemia and ischemic tolerance produced by ischemic preconditioning⁸⁰. In the ipsilateral cortex, ischemia altered interaction of PPP1α and γ with overlapping and distinct proteins. While both isoforms exhibited altered binding to HSC70 and DARPP32, changes in association with PDE6B, VCP, GRP78, neurofilament light polypeptide, DRP2, prelamin-A/C, γ-enolase, β-actin, and SIAH2 were specific to PPP1γ. Preconditioning restored some ischemic changes to control levels, while newly affecting others⁸⁰. GRP78, neurofilament light polypeptide, and DARPP32 were known regulatory components of PPP1⁸¹, whereas PDE6B, creatine kinase-β, and SIAH2 were novel interactors that remain unexplored. Only VCP binding to PPP1 has been independently confirmed⁸².

Cerebral ischemia causes translational arrest through binding of 4E-BP2 to eIF4E, inhibiting its function⁸³. To reveal other potentially important proteins in this regulation, 4E-BP2-interacting proteins were identified by proteomics in the ischemia-resistant cortex and vulnerable hippocampus in sham and BCCAO animals⁸⁴. Interestingly, the 4E-BP2 interaction profile in the ischemia-resistant cortex compared to the susceptible hippocampus was already vastly different at baseline, which could explain differing vulnerabilities of these areas. Of note, HSC70, DRP2, α-enolase, UCH-L1, adenosine kinase-1, GAPDH, and phosphoglycerate kinase-1, maintained their region-specific interaction with 4E-BP2 after ischemia⁸⁴.

While proteomic association studies led to the discovery of many potential new regulators of PKC, PPP1, and 4E-BP2, follow-up studies on the functional implications of modified interactions in the context of ischemia and preconditioning are lacking.

2.1.5.2. PTMs and protein aggregation

In the acute phase, cerebral ischemia activates nuclear small-ubiquitin-like modifier (SUMO)2/3 conjugation, which is associated with neuroprotection^85–87. To uncover molecular mechanisms that link nuclear SUMOylation to protection, targeted proteins were identified in the ischemia-resistant cortex after BCCAO and 1h reperfusion in SUMO-transgenic mice, using a SUMO-enrichment and proteomics approach⁸⁸. Ischemia induced SUMOylation of 91 proteins, among them many RNA (heterogenous nuclear ribonucleoproteins, RBM14, DEAD-box helicase-5, and DExH-box helicase-9) and transcriptional (glucocorticoid receptor, TFII-I, TIF1B, and CTIP2) regulators, supporting a critical role for transcriptional and translational regulation in SUMO-mediated neuroprotection.

Like SUMOylation, ubiquitin conjugation is activated in the acute phase of ischemia^89,90. However, in contrast to SUMO, the significance of ubiquitination in post-ischemic cell fate remains undetermined. To assess potential functions of ubiquitination, the ubiquitinated proteome was analyzed by ubiquitin-enrichment and subsequent proteomics in hippocampi of mice after BCCAO and 4h reperfusion⁹¹. Compared to sham, 272 proteins exhibited increased ubiquitination, including many synaptic regulators such as CAMK2A, PKC, creatine kinase-β, GRIN2B, DLGAP1, NSF, and SynGAP1, likely impacting synaptic function. Additionally, some translation factors (eIF4H and eEFA1) and chaperones (HSPA12A and DNAJA1) were highly ubiquitinated, potentially regulating the reported translational arrest after ischemia³.

Ubiquitin attaches to substrates as monomers or chains, and the conjugation type determines the impact of ubiquitin modification on protein function^92,93. Proteomics found that ischemia induced ubiquitination via lysine 6-, 11-, 48-, and 63-conjugated chains⁹¹, indicating multiple involved mechanisms and regulation of a wide range of cellular processes. Notably, post-ischemic ubiquitinated proteins are predominantly detergent-insoluble^89,91, suggesting sequestration into potentially toxic aggregates. This has not yet been verified.

Ischemia affects solubility of proteins in the acute phase, as found by proteomics of detergent-insoluble material obtained from cortices of mice after MCAO and 1hr reperfusion⁹⁴. After ischemia, 196 proteins exhibited increased insolubility, the largest group being associated with RNA processing, including TDP43, FUS, HNRNPA1, NONO, and SFPQ. Accumulation of such proteins is commonly linked to neurodegenerative diseases (NDs), where it is terminal⁹⁴. However, it is transient after ischemia if reperfusion occurs⁹⁴. This indicates a potential functional role for RNA-binding protein sequestration following ischemia, instead of foreshadowing cell death. Other proteins with increased insolubility after ischemia were HSPs, ubiquitin, SUMO, eIF4, and kinases PKC and CAMK2. The mechanism by which these proteins become insoluble awaits to be established but could be related to protein aggregation or simple change in subcellular localization.

2.2. Human studies – IS

To date, several proteomics studies, published mostly by the Montaner group, have identified approximately 190 DEPs in postmortem human brains at different stroke stages (Table 1 and Table S4). For data interpretation, effects of high patient heterogeneity and potential changes in pathways after death must be considered. First, DEPs were assessed by global proteomics in the contralateral hemisphere, peri-infarct area, and infarct core of postmortem brains from three patients with MCAO and compared to three controls without any apparent neurologic condition^95,96. In the infarcted tissue, 51 DEPs were identified. GFAP and albumin were increased in infarct and peri-infarct areas, likely reflecting astrogliosis and trapped blood or blood-brain barrier (BBB) damage, respectively. Other DEPs of interest include plasma proteins (C3, fibrinogens, and transferrin), vesicle transport proteins (NSF, GDI, and ARHGDIA), HSC70, gelsolin, and DRP2, all of which were also altered in infarcted animal brain tissue after MCAO^{7,8,11–16,23,24,30,32,36,38}. Interestingly, gelsolin, cystatin-A, and DRP2 were changed not only in the brain parenchyma, but also in blood from stroke patients, where altered levels were associated with long-term outcome⁹⁶.

Table 1.

Proposed biomarkers for stroke diagnosis and prognosis discovered by proteomics in patients. A comprehensive list of identified DEPs from each study can be found in Tables S4 (tissue) and S6 (biofluids).

Samples	Analysis cohort (n)	Sampling time	Methods	Major findings	Samples	Blood biomarker utility	Parameters (Cutoff, sensitivity, specificity)	Ref.
PROTEOMICS DISCOVERY					BIOMARKER EVALUATION
Blood					Blood
Plasma healthy, IS, HS	healthy (n=21), IS (n=11), HS (n=10)	At admission	Protein chip, Gel-based MS	DEPs: ApoC-I, ApoC-III, SAA, AT-III (IS vs HS)	6h after symptom onset, healthy (n=12), IS (n=16), HS (n=15)	Diagnosis: distinguish between IS and HS	ApoC-I (RFU, 94%, 73%), ApoC-III (RFU, 94%, 87%)	140
Serum TIA, minor IS, mimics	TIA (n=20), minor IS (n=15), mimics (n=12)	Within 48h of symptom onset	TMT, LC-MS/MS	DEPs: CP, C8γ, PBP (TIA/minor IS vs mimics)	Within 48h of symptom onset, TIA (n=22), minor IS (n=20), mimics (n=14)	Diagnosis: distinguish between TIA/minor IS and mimics	PBP (1.5μg/ml, 91%, 57%)	137
Serum healthy, IS	pooled samples each n=20	Within 24h of symptom onset	iTRAQ, LC-MS/MS	60 DEPs (IS vs healthy)	Within 24h of symptom onset; IS (n=50), healthy (n=35)	Diagnosis: distinguish between IS and healthy	↑ vWF (N/A, 90%, 83%), ↓ ADAMTS13 (N/A, 98%, 90%), ↓ S100A7 (N/A, 97%, 91%)	132
Serum TIA, minor IS, mimics	TIA/minor IS suspected (n=545)	Within 24h of symptom onset	LC/MRM-MS	16 DEPs (TIA/minor IS vs mimics)	Within 24h of symptom onset; TIA/minor IS (n=811), mimics (n=292)	Diagnosis: distinguish between TIA/minor IS and mimics	IGFBP-3, PON3 are promising TIA predictors but failed to achieve clinical relevance	138,139
Whole blood IS, HS, mimics	IS (n=19), HS (n=17), mimics (n=20)	At admission	Peptide array	Altered circulating antibody pool (IS/HS vs mimics)	Same cohort	Diagnosis: distinguish between IS/HS and mimics; between IS and HS	Model with 17 peptide probes: IS/HS vs mimics (N/A, 92%, 90%); IS vs HS (N/A, 93%, 90%)	134
Serum IS	IS (n=21)	5 timepoints from imaging to 90d after IS	SELDI-TOF-MS/MS	23 DEPs with significant change over time	Correlation analysis performed with proteomics data	Prognosis: vWF is associated with poor outcome (3mth)	N/D	145
Plasma healthy, IS with LVO	healthy (n=20), LVO IS (n=40)	Within 7d of symptom onset	iTRAQ, LC-MS/MS	DEPs: THBS1, LYVE1, IGF2, PPBP, SPP2, CDH1, APOC4-APOC2	Within 7d of symptom onset, healthy, non-LVO IS, LVO IS (n=33/group)	Prognosis: ↑ IGF2, LYVE1, ↓ THBS1 are associated with favorable outcome (3mth)	N/D	131
Plasma IS with LVO	IS with LVO (n=40)	At the time of thrombectomy	Antibody-based	DEP: VCAM1	Correlation analysis performed with proteomics data	Prognosis: ↑ VCAM1 is associated with poor outcome on discharge	N/D	157
Serum IS w/o reperfusion therapy	HT (n=5), non-HT (n=32)	At admission	iTRAQ, LC-MS/MS	17 DEPs (e.g., A1M, FBLN1) (HT vs non-HT)	Correlation analysis performed with proteomics data	Prognosis: DEPs alone or in combination could predict HT	N/D	143
HDL healthy, IS	healthy (n=35), IS 24 and 96h post intervention (n=35, n=20)	24 and 96h post intervention	PRM-MS/MS	12 DEPs (e.g., APOE, CLU, SAA2, APOF) (IS vs healthy)	Correlation analysis performed with proteomics data	Prognosis: APOF, APOL1 levels correlate with recovery (3mth)	N/D	147
Other biomaterials					Blood
CSF healthy, deceased	healthy, deceased (each n=4)	Unspecified	2-DE-MS	DEP: H-FABP (decreased vs healthy)	40min to 3d after symptom onset, healthy (n=22), MI (n=20), IS (n=22)	Diagnosis: distinguish between stroke and healthy controls	H-FABP (OD405=0.531, 68.2%, 100%)	123
ECF IS with malignant MCA infarction	IS (n=6)	Within 24h after reperfusion	TMT, MALDI-TOF/TOF MS	53 DEPs (e.g., UCH-L1, APOA4, UBIQ, AQP4)	Within 24h of symptom onset, healthy (n=14), IS (n=12), HS (n=2)	Diagnosis: ↑ GSTP1, S100B, PRDX1 distinguish between stroke and healthy controls	N/D	125
Thrombi IS with LVO	IS with LVO treated by mechanical thrombectomy (n=41)	210–344min after IS	LC-MS/MS	DEPs: glycophorin-A, fibrinogen (cardioembolic vs atherothrombotic)	ESUS (n=12)	Diagnosis: distinguish between cardioembolic and atherothrombotic IS	Model with 40 parameters selected from 2456 proteomics and 5019 metabolomics features (N/A, 100%, 85.7%)	149
Tissue (infarct, peri-infarct) healthy, IS	healthy, IS (each n=3)	6h postmortem	2-DE-MS	51 DEPs (e.g., FGG, CSTA, LDHA) (infarct core vs peri-infarct)	Within 4.5h of symptom onset, IS (n=60)	Prognosis: combination of ↑ GELS, CYTA, ↓ DRP2 predicts poor long-term outcome (3mth)	GELS (19.87ng/mL, 71.8%, 76.5%), CYTA (31.76ng/mL, 26.8%, 100%), DRP2 (2.09ng/mL, 97.1%, 28.6%)	96
Neurons + BBB (infarct, contralateral) IS	IS (n=7)	4–9h postmortem	LC-ESI-MS/MS	30 DEPs (e.g., SH3G2, EAA2, SRSF1, SAHH2) (infarct vs contralateral)	Within 6h of symptom onset, healthy (n=8), mimics (n=13), IS (n=45)	Prognosis: ↓ SAHH2 predicts better neurologic outcome at 24 and 48h after IS	SAHH2 (993.23pg/mL, 89%, 58%)	97

Abbreviations: CSF, cerebrospinal fluid; DEP, differentially expressed protein; ECF, extracellular fluid; ESUS, embolic stroke of undetermined source; HS, hemorrhagic stroke; IS, ischemic stroke; LVO, large vessel occlusion; MCA, middle cerebral artery; MI, myocardial infarction; N/D, not determined; RFU, relative fluorescence unit; TIA, transient ischemic attack.

Subsequently, laser microdissection was combined with label-free MS proteomics to analyze cell-specific proteomes of neurons and BBB structures isolated from the infarct area and contralateral regions of human IS brains⁹⁷. This analysis identified 30 DEPs with little overlap to previous studies, including SAHH2, PACN1, and SRSF1 in neurons, and neuromodulin, PKCγ and SIRT2 in the BBB. The distinct DEPs identified in this study likely derive from different starting material, comparison group, and/or proteomics approach used in previous studies. In a cohort of patients, the authors found that lower circulating levels of SAHH2 at admission was an independent predictor of neurologic improvement after 24h⁹⁷. However, there was no follow-up stroke biomarker study on SAHH2.

Recently, a combined proteomics and transcriptomics approach was implemented to identify DEPs in the infarct core vs contralateral hemisphere of six human postmortem stroke brains⁹⁸. Only 60% of DEPs were equally regulated at both mRNA and protein levels, emphasizing the necessity of proteomics in the search for stroke therapeutic targets and biomarkers. Among the 66 DEPs, Ras-related protein-3C and sulfotransferase-4A1 were confirmed to be down-, and α−1-acid glycoprotein 1 and α−1-antitrypsin up-regulated in another cohort of brain samples. The increase in brain α−1-acid glycoprotein 1 and α−1-antitrypsin levels translated to a significant elevation of both proteins in blood from 11 acute IS patients treated with t-PA within the first 24h after stroke onset. Further exploration of these two proteins as biomarkers is needed.

Finally, a proteomics study not published by the Montaner group, used 8-plex iTRAQ labeling to quantitatively determine DEPs in the putamen, thalamus, and parietal lobe of postmortem human stroke brains⁹⁹. Compared to location-matched controls, several proteins were increased in infarcted tissue, including vimentin, annexin-2, HSP70, MBP, and ferritin. All these DEPs were also altered in MCAO-treated animals^4,5,23,32,71, indicating their involvement in stroke pathogenesis across both species. Additionally, all identified mitochondrial DEPs (e.g., SLC25A11, GOT2, and NDUFS1) were decreased in the infarcted tissue, signifying dysregulated mitochondrial function.

3. Proteomics assessment of biofluids: in search of stroke biomarkers

Stroke protein biomarkers would aid in stroke diagnosis, prognosis, and therapy efficacy evaluation (Figure 1). However, while protein biomarkers are routinely used in the clinic for some cardiovascular diseases (e.g., natriuretic peptides for heart failure), stroke biomarker discovery has not yet come to fruition. Most large clinical studies selected biomarker candidates based on current knowledge of stroke pathophysiology, which has had limited success. This was underscored by the Stroke-Chip study, which found that 21 commonly investigated biomarkers, including HSC70, NT-proBNP, matrix metalloproteinase-9, D-dimer, vWF, γ-enolase, S100B, interleukin-6, cellular fibronectin, and apolipoprotein-CIII, were unable to attain sufficient diagnostic accuracy in discriminating stroke from stroke mimics, even after combining with clinical variables¹⁰⁰. Further, stroke subtype discrimination has been the subject of many clinical studies focusing on traditional biomarkers, of which GFAP currently is considered the most promising with excellent specificity to differentiate ICH (but not SAH) from IS^101–103. However, its sensitivity to rule out ICH for thrombolytic therapy is a concern, and accompanying biomarkers are still needed. Some commonly studied biomarkers such as C-reactive protein¹⁰⁴, neurofilament light polypeptide^105,106, and C3¹⁰⁷, have some predictive value for stroke outcome. None, however, acquired sufficient high-quality evidence with adequate performance to justify their clinical use. For recent general reviews on blood stroke biomarkers, see^101,108.

Proteomics to assess post-stroke protein changes in body fluids, such as CSF, extracellular fluid (ECF), urine and blood, creates an advanced opportunity to identify novel biomarkers that are superior to the current candidates. While collecting CSF and ECF from patients is an invasive procedure, these fluids are attractive for biomarker discovery due to their proximity to the diseased brain. Blood is considered the best source for stroke biomarker discovery, as it is easily and non-invasively collected, and contains brain-specific, stroke-related proteins released from (dying) brain cells that enter the circulation once BBB integrity deteriorates. Urine is easy to obtain as well, contains a less complex proteome than CSF or blood, and has proven valuable for diagnosing heart disease and nephropathy¹⁰⁹. Its utility in stroke diagnosis, however, is unclear.

More recently, stroke biomarker discovery has also been attempted in patient-derived blood clots and extracellular vesicles (EVs). With increased use of endovascular thrombectomy to treat stroke patients with large vessel occlusion (LVO), thrombi can be retrieved from living patients. This unique tissue sample could give essential information about stroke-related thrombogenesis and the origin of blood clot formation to diagnose IS subtypes (thrombotic vs embolic). EVs are present in various biofluids including blood, and carry proteins, nucleic acids, lipids, and metabolites from parent cells, thus providing information about diseased tissue as found in stroke¹¹⁰. Brain-derived EVs cross the BBB, and can be recovered from peripheral blood to analyze cargo.

In the following, we discuss proteomics studies for biomarker discovery in biofluids, blood clots, and EVs from stroked animals and human stroke patients. See Tables S1 and 1 for a complete list of animal and human studies, respectively, and Tables S5 and S6 for a listing of identified DEPs.

3.1. Animal studies

IS.

Only a few animal studies have examined post-stroke proteomic changes in biofluids from rodents and nonhuman primates (Table S1). While most of the ~900 identified DEPs require further assessment in humans to be considered potential biomarkers, some have been evaluated in stroke patients. One protein that was found altered in several rodent proteomics studies is transthyretin. Levels of transthyretin were increased in CSF¹¹¹ and urine¹¹², but decreased in plasma¹¹³ within 8–24h of MCAO or spontaneous IS. Low serum transthyretin levels were also observed in stroke patients within a 24h time window¹¹⁴. In addition to transthyretin, proteomics consistently identified DEPs connected to the inflammatory and coagulation response in blood^112,113,115, urine^112,115, and CSF^112,115,116 after IS in rodents and nonhuman primates. Altered proteins in blood included C3, α−2-macroglobulin, haptoglobin, albumin, and transferrin, which were also modified in serum from stroke patients, where they may serve as prognostic tools for stroke outcome^{46,107,117–121}. In the CSF, DEPs again included C3, α−2-macroglobulin, haptoglobin, and transferrin. Proteins related to immunosuppression and M1/M2 macrophage transition (adiponectin, coagulation factor XIII, α−1-acid glycoprotein 1, MBL2, and stabilin-1), angiogenesis (multimerin-2) and immunomodulation (immunoglobulin J-chain and IGLL1) were increased in CSF by pretreatment with the neuroprotective toll-like receptor (TLR)9 agonist D192935, potentially reflecting its therapeutic effects on the brain¹¹⁶.

Exclusively relying on inflammatory serum DEPs as stroke biomarkers in the clinic may be problematic due to their variability in response to many other conditions that affect the immune system. Therefore, it is important to include other proteins on a potential stroke biomarker panel. Using an aptamer-based proteomics approach, Simats et al identified 716 DEPs in rat CSF during the hyperacute stroke phase¹²². Among the top elevated proteins were many kinases such as creatine kinase-β, CAMK2A, CAMK2B, and CMPK. Creatine kinase-β and CMPK were also elevated in blood from human stroke patients within 6h of symptom onset, which could be useful for stroke diagnosis. While CAMK2B could not distinguish between stroke and control subjects, high plasma levels predicted poor outcomes in a cohort of stroke patients, as did CMPK¹²².

Although most identified inflammatory DEPs in CSF and plasma are extracellular by nature, some, such as kinases identified in this study, are normally localized intracellularly, and are likely derived from dying cells. By comparative post-analysis of animal tissue and fluid proteomics data (see Table S3, sheet 3, for a list), we found that increased presence of CAMK2 and CMPK in CSF and blood was accompanied by decreased tissue levels (Figure 4). The same was true for other common intracellular proteins found in fluids after stroke (γ-enolase, calmodulin-1, and UCH-L1). In contrast, elevation of “classic” plasma proteins, such as α−2-HS-glycoprotein, C3, and transthyretin, in post-stroke fluids correlated with an increase in tissue levels, likely indicating extravasation into the brain parenchyma (Figure 4).

Figure 4.

Venn Diagrams illustrating common DEPs after MCAO in animal brain tissue and CSF. (A) 20 DEPs in blue were upregulated in both tissue and CSF. They largely constitute “classic” plasma proteins. (B) 22 DEPs shaded in brown were decreased in tissue, while increased in CSF. They are mainly intracellular proteins. There is no overlap of proteins downregulated in CSF with any proteins detected in tissue. All DEPs used for this analysis are listed in Table S3. CSF, cerebrospinal fluid.

3.2. Human studies

3.2.1. CSF

IS and HS.

Proteomics comparing CSF between deceased and healthy subjects identified H-FABP as a marker of brain necrosis, which could be useful in diagnosing stroke with associated necrotic cell death¹²³. Using quantitative measurement of H-FABP levels via ELISA in serum from healthy controls, as well as patients with hemorrhage, IS, and acute myocardial infarction, the authors were able to discriminate between the groups¹²³. The specificity and sensitivity of serum H-FABP for stroke diagnosis were higher than for γ-enolase and S100B, two proteins that have been extensively assessed for this purpose¹⁰¹. In another study, targeted proteomics showed that γ-enolase and S100B, alongside five other proteins (GFAP, MBP, neurofilament medium polypeptide, α-internexin, and β-synuclein), were elevated in CSF from HS patients, where they were significantly higher than in controls and patients with IS¹²⁴. Neurofilament medium polypeptide, α-internexin, and β-synuclein have not yet been evaluated as stroke biomarkers.

3.2.2. ECF

IS.

Proteomic analysis of ECF microdialysates from IS patients with malignant MCA infarction found 53 proteins that were increased in the ischemic penumbra and core compared to the unaffected contralateral side¹²⁵. Several of these proteins, e.g., FABP, GFAP, S100B, and MBP, were also increased in CSF and blood after stroke^123,124. Three biomarker candidate proteins, GSTP1, peroxiredoxin-1, and S100B, were confirmed elevated in blood in a cohort of stroke patients (n=14).

SAH and ICH.

Delayed cerebral vasospasm is a critical pathologic determinator of SAH outcome¹²⁶, and identifying associated biomarkers would potentially aid in early clinical intervention. The ECF proteome from vasospastic and non-vasospastic SAH patients was compared 1.5d before the onset of symptomatic vasospasm¹²⁷. Differential levels of GAPDH and HSP7C were detected, which, with validation, could be biomarkers for predicting development of symptomatic vasospasm after SAH.

Proteomics of ECF microdialysates collected from the peri-hemorrhagic zone and healthy cortex of ICH patients at time of surgery was performed to identify potential biomarkers that predict extent of secondary injury¹²⁸. A few DEPs, including increased hemoglobin and transthyretin, and decreased haptoglobin and IGHA1, were identified, but no validation was performed. Consequently, the suitability of any of these proteins as ICH biomarkers has not yet been confirmed.

3.2.3. Urine

IS.

The proteome of urinary samples collected from stroke patients within 24h of symptom onset was compared to healthy controls¹²⁹. Hemoglobin, ITIH4, β−2-microglobulin, ceruloplasmin, GGT2P, and collagen α−1(II) chain were among the DEPs identified; however, only the latter two passed rigorous adjustment for multiple testing as stand-alone biomarkers for discriminating between groups. To increase performance accuracy, the authors combined 35 DEPs to a biomarker panel, which enabled stroke diagnosis in a small patient cohort with good confidence. This panel has not yet been tested for capacity to distinguish between stroke and mimics.

3.2.4. Blood

3.2.4.1. Stroke diagnosis

Differentiating between IS patients and healthy controls.

An early proteomics study found significantly increased serum hemoglobin levels in acute IS patients within 3d of stroke onset compared to controls¹³⁰, introducing it as a potential stroke biomarker. However, blood collection in this study occurred at a timepoint far beyond the current therapeutic window for IS (24h).

iTRAQ-based LC-MS/MS was used to assess proteomic changes in plasma from IS patients with LVO¹³¹. A panel of four elevated proteins, platelet basic protein, thrombospondin-1, LYVE1, and IGF2, after stroke was identified and, when used on a different patient cohort, was able to discern stroke LVO patients from controls. Further, IGF2, LYVE1, and thrombospondin-1 proved to be independent predictors of good outcome at a 3mth follow-up. Although this is a promising result, the blood collection timepoint of up to 7d after stroke onset, again makes the panel impractical from a therapeutic standpoint. The same approach with blood samples from IS patients within 24h after symptom onset led to discovery of 66 DEPs, many of them with known impact on stroke pathology¹³². Compared to healthy controls, C-reactive protein, β−2-microglobulin, lipopolysaccharide-binding protein, C3, ADAMTS13, vWF, and thrombospondin-4 were increased, while S100A7 and platelet basic protein, in contrast to the prior study, were decreased. After vWF, ADAMTS13, and S100A7 serum levels were further evaluated by ELISA, they showed excellent diagnostic value. However, validation analysis using a different cohort is needed.

Distinguishing stroke from stroke mimics.

One challenge of stroke diagnosis is distinguishing strokes from stroke mimics, other medical conditions with clinical symptoms similar to stroke, such as seizures, migraine, and delirium. Using targeted proteomics, 147 proteins were assessed for their potential to differentiate between stroke patients and patients with stroke-mimicking conditions¹³³. Thirty DEPs were identified, and used to generate a model that exhibited high diagnostic value. Rapidly and simultaneously measuring 30 proteins, however, is not practical in the current clinical setting. Another study used a high-density peptide array to determine whether altering binding characteristics of circulating antibodies in plasma could be used for stroke diagnosis¹³⁴. Combined antibody-binding intensities to 17 peptide probes demonstrated encouraging sensitivity and specificity to discriminate stroke from stroke mimics, indicating potential for antibody-based stroke diagnostics.

Discriminating between IS and TIA patients.

TIA is a risk factor for subsequent stroke that can be mitigated with appropriate prevention¹³⁵. It remains difficult, however, to diagnose TIA patients, with once promising biomarkers failing to improve clinical practice¹³⁶. One study used global quantitative proteomics to analyze serum samples from patients presenting with TIA, minor stroke, and mimicking conditions¹³⁷. PBP was increased after stroke, which was further validated by ELISA in a second cohort of patients. While elevated PBP levels diagnosed stroke, they were unable to separate TIA from minor stroke patients. More comprehensive studies on biomarker-based distinction of minor stroke and TIA patients were recently performed by the SpecTRA study group^138,139. In phase 1, targeted proteomics of plasma from patients within 24h of symptom onset led to the development of a 16-DEP panel, from which nine proteins were also significant univariate predictors of TIA vs mimics¹³⁸. In phase 2, IGFBP3 and PON3 were validated as TIA biomarkers in two additional patient cohorts¹³⁹, which if independently confirmed, would be of great clinical value.

Distinguishing IS from HS.

Having a reliable blood biomarker to facilitate early distinction between these two stroke types would greatly improve treatment selection. Plasma proteomics on samples obtained from healthy controls, and IS and HS patients at admission, showed higher levels of apolipoprotein-C-I and apolipoprotein-C-III in IS, which was further verified by ELISA¹⁴⁰. Encouragingly, apolipoprotein-C-I and apolipoprotein-C-III were confirmed as suitable biomarkers in an independent study using targeted proteomics to measure apolipoprotein levels in plasma samples from stroke patients¹⁴¹. This study further indicated that the ischemic vs hemorrhagic groups were best differentiated by combining measurements of apolipoprotein-C-III and apolipoprotein-A1. Collectively, apolipoproteins make good biomarker candidates.

A recent proteomics study analyzed plasma samples from IS and ICH patients, and healthy controls, and identified nine biomarker candidates. Three of them, with higher levels in IS (CPN2, coagulation factor XII, and MASP1), were related to the blood coagulation system, and showed potential for differentiating between IS and ICH¹⁴².

3.2.4.2. Stroke prognosis

A few proteomics studies have aimed to identify prognostic blood biomarkers that predict functional recovery after stroke. The development of hemorrhagic transformation (HT) is a severe complication following IS that exacerbates brain damage. To optimize treatment of stroke patients at high risk for developing HT, a blood biomarker that predicts HT is of urgent clinical interest. The first systematic approach to identify such biomarkers was only recently performed using quantitative proteomics¹⁴³. Analysis of serum samples taken from a small group of IS patients with no recanalization, some of which spontaneously developed HT (5 out of 37), identified 17 DEPs, including apolipoprotein C-II, HSP72, and CRHBP. Future studies are required to assess their predictive values.

Longitudinal quantitative proteomics of plasma collected from IS patients at three time points post stroke (days up to 12mth), was performed to track protein expression changes during recovery¹⁴⁴. Several DEPs were identified, many associated with the complement pathway. However, none were independently evaluated as biomarkers. Another longitudinal proteomics study assessed serum samples from a cohort of IS patients treated with placebo or t-PA at five timepoints from baseline imaging to 90d after stroke¹⁴⁵. The blood protein composition after stroke was highly dynamic, indicating the importance of timing blood sample collection for biomarker discovery. Twenty-six DEPs were altered with t-PA treatment, while 23 changed over the course of stroke progression. Baseline vWF levels correlated with functional recovery at 3mth after IS. Disappointingly, no other DEP was significantly associated with stroke subtype (atherosclerotic vs cardioembolic), stroke severity, HT, or cardiovascular risk factors.

High-density lipoprotein-cholesterol (HDL-C) levels are routinely used as an inverse risk biomarker for cardiovascular disease¹⁴⁶, but little is known about changes in the HDL protein cargo after IS. A recent targeted proteomics study investigated the potential association between HDL proteome composition and functional recovery after IS¹⁴⁷. Changes in nine proteins (apolipoproteins E, F, L1, M, and C-IV, as well as APMAP, LPA1, prenylcysteine oxidase-1, and PON1) significantly correlated with recovery scores at 3mth post stroke. Thus, these proteins may be useful as biomarkers to predict stroke recovery. With this encouraging data, further validation in a different cohort of stroke patients is strongly warranted.

3.2.5. Blood clots

To discover proteins suitable for use as biomarkers of stroke etiology, an exploratory proteomics study obtained thrombi from 20 patients with acute IS, and analyzed for DEPs that correlated with clinical features¹⁴⁸. The levels of three DEPs (septin-2, phosphoglycerate kinase-1, and integrin α-M) correlated with serum low-density lipoprotein (LDL), while one protein, septin-7, corresponded to the erythrocyte sedimentation rate. However, due to the small sample size, only limited conclusions can be drawn from this study.

Analyzing thrombi from a larger cohort of patients (34 cardioembolic and 7 atherothrombotic), a recent omics study identified a distinct proteomic signature in the two etiologies, with higher glycophorin-A (indicative of more red blood cells) and fibrinogen in cardioembolic stroke patients¹⁴⁹. Combining proteomic and metabolomic data offered an excellent predictive model for distinguishing the two stroke subtypes with a diagnostic performance of 100% sensitivity/85.7% specificity. However, the model uses 40 omics features, which is technically not feasible for current clinical use. Given the great diagnostic value of thrombi for stroke etiology, further research in this new field is needed.

3.2.6. EVs

Proteomic changes in circulating EVs following two different cardiovascular conditions were assessed by comparing serum EV content in 81 IS patients, 37 myocardial infarction patients, and 22 healthy controls¹⁵⁰. Retrieval of EVs occurred within 24h and 72h of symptom onset after brain and heart infarction, respectively. EVs contained an average of 146 proteins, out of which only three, apolipoprotein-B, α−2-macroglobulin, and fibronectin, showed altered levels between ischemic and myocardial infarction. Disappointingly, no significant change in EV protein content was found between IS and control subjects. However, after separating the same stroke patient group into lesion subtypes (26 subcortical and 55 cortical-subcortical), LC-MS/MS analysis identified some proteins in EVs that were unique to one subtype, e.g., complement C1QA and caspase-14 in cortical-subcortical, and annexin-2 in subcortical lesions¹⁵¹. Compared to controls, EVs from cortical-subcortical patients had significantly elevated FGC, fibronectin, and vWF levels, indicating heightened atherosclerosis. In addition to investigating EV content by proteomics, post-stroke changes in EV cell populations based on cell origins (e.g., endothelial cells and platelets), are currently under investigation as prognostic biomarkers^152,153.

4. Proteomics from biofluids: a reverse translational approach to identify mechanisms and targets

A few clinical proteomics studies followed a reverse translational approach to gain insight into pathologic mechanisms and therapeutic targets through DEP analysis of blood from stroke patients. Cervical artery dissection (CAD) is a leading cause of stroke in young adults. To understand the underlying pathophysiologic mechanisms, which are largely unknown, one study quantitatively compared the serum proteome between CAD stroke and non-CAD stroke patients from blood collected up to 10d from symptom onset¹⁵⁴. Lipopolysaccharide-binding protein, VCAM1, ficolin-2, fibulin-1, apolipoprotein-C-I, and apolipoprotein-B were upregulated after CAD stroke, pointing to a difference in immune response, blood coagulation, and lipid metabolism.

To understand sex differences in stroke pathophysiology, comparative proteomics was used to analyze DEPs in plasma collected from male and female IS patients (mean age = ~82 years) prior to thrombolysis and within 24h of symptom onset¹⁵⁵. Expression of 77 proteins differed between the two sexes, with higher levels of most DEPs in women. CBG, a negative regulator of cortisol bioactivity, was 16 times higher in females vs males, indicating involvement of cortisol signaling in the sexually dimorphic outcome. Interestingly, some of the identified DEPs with elevated levels in females are proteins involved in various stages of thrombosis, such as fibrinolysis, coagulation, and platelet activation, potentially underlying the higher risk for stroke in older females.

The increased use of thrombectomy enables collection of intracranial blood. Comparison of intracranial and systemic blood protein content can provide insights into the pathophysiologic changes in the stroke brain¹⁵⁶. In a pilot study, plasma proteome analysis in patients showed that all DEPs, most significantly ficolin-2, fetuin-B, FAP, uromodulin, and PLTP, exhibit lower intracranial levels compared to systemic arterial blood. Whether these DEPs are stroke-specific or simply reflect basic differences between blood sources remains to be clarified. In a follow-up study, the authors showed that intracranial, but not systemic, levels of VCAM1 were positively associated with infarct and edema volume, suggesting a detrimental role of VCAM1 in stroke pathology¹⁵⁷.

5. Challenges and future perspectives

Stroke proteomics, with its capability to analyze protein changes in a biological sample on a large scale, is a powerful approach that could advance stroke diagnosis and intervention by facilitating and expediting discovery of biomarkers and novel therapeutic targets. Since the birth of stroke proteomics, many intriguing DEPs have been identified in brain tissue as well as biofluids from stroke patients and animal models. None of these proteins, however, has led to a clinically approved stroke biomarker or therapy. The reasons for this translational gap are many.

Therapeutic target discovery by stroke proteomics is complicated by many factors. First, there is a relatively poor overlap of DEPs among current studies, which is likely due to the different experimental conditions used, particularly concerning stroke models, timing of sampling, and methods of protein preparation and proteomics. Second, there is a lack of follow-up studies confirming key findings that are truly related to stroke pathophysiology. While proteomics is highly effective in identifying proteins, it cannot inform the role of a DEP in stroke pathology or determine whether the altered protein level is a cause or consequence of injury. Third, inaccurate data interpretation may hinder discovery. As the name “DEP” implies, altered protein levels detected in tissue are commonly interpreted as a change in protein expression, which may not always be the case. For instance, many studies use little or no detergents during tissue extraction. This precludes analysis of insoluble cellular material such as the plasma membrane, the postsynapse, or aggregates, to which proteins could have translocated. Another reason for elusive therapeutic targets might be that they are not found by assessing protein level changes, but rather by evaluating changes in protein modifications after stroke, which are vastly understudied. Finally, all current experimental stroke studies were executed in young animals, which is an obvious mismatch to the clinic, where most stroke patients are elderly (>60 years), and functional outcome is inversely associated with age¹⁵⁸. Thus, proteomics analysis of aged, stroked animals is urgently needed.

Biomarker discovery by stroke proteomics has been hampered by lack of candidates that consistently perform between different studies and lack of validation in large cohorts of heterogeneous stroke patients with sufficient ethnic and geographic diversity. Rather than relying on individual labs, it may be more productive to fund multi-lab or -center initiatives that ensure strict patient selection, standardized sample collection, and use of state-of-the-art proteomics technology and data analysis for biomarker discovery. Another strategy to improve biomarker performance is using a biomarker protein panel that covers different aspects of stroke pathophysiology, such as thrombosis, inflammation, and brain injury. Ideally, this panel should be feasible for clinical use and include stroke-specific brain-derived proteins, which are likely present in blood at low levels (pg-ng/mL). Identifying such proteins via proteomics presents a technical challenge, as the human blood has a highly complex and dynamic proteome¹⁵⁹. Increasing sensitivity of mass spectrometers, and depleting high-abundance proteins or enriching stroke-specific proteins by affinity-purification with antibodies, for example, may help solve this problem. Alternatively, some proteomics studies have assessed protein changes in urine, which has a less complex proteome than blood.

Some newer approaches in stroke proteomics take advantage of enriched brain-specific proteins in EVs and thrombi. Several brain-cell types, including neurons and endothelial cells, release EVs into the blood during stroke. Analyzing their contents may provide new insights into stroke pathophysiology, and inform biomarker development. Current efforts are directed at optimizing EV isolation¹⁶⁰. Additionally, while clot proteomics is still in its infancy, and needs further optimization, stroke centers have started to recognize the potential value of using thrombi for stroke diagnosis¹⁶¹, which will help move this field forward.

As omics technologies enter clinical research, combining analysis of complex biological data with multi-dimensional patient information requires a systems approach using the highest level statistical and bioinformatic methodologies^162,163. Evolving methods of artificial intelligence and machine learning are well-suited for this purpose, and are beginning to be adopted in biomarker discovery^164,165. It is with great hope that combining (prote)omics with these technologies will lead to a breakthrough in identifying biomarkers and therapeutic targets, and thereby revolutionize stroke diagnosis and care.

Abbreviations not described in the text (in alphabetical order)

4E-BP2

eukaryotic translation initiation factor 4E-binding protein-2

ACO1

cytoplasmic aconitate hydratase

ACO2

aconitate hydratase mitochondrial-2

ADAMTS13

a disintegrin and metalloproteinase with thrombospondin motifs-13

ALDOC

fructose-biphosphate aldolase-C

APMAP

adipocyte plasma membrane-associated protein

ARHGDIA

Rho GDP-dissociation inhibitor-1

ATP1A1

sodium/potassium-transporting ATPase subunit α1

ATP5C1

ATP synthase subunit γ, mitochondrial

ATP6V1F

V-type proton ATPase subunit-1F

BAIAP2L1

brain-specific angiogenesis inhibitor 1 associated protein-2L1

BASP1

brain acid soluble protein-1

CAMK2A

calcium-calmodulin-dependent protein kinase type II subunit α

CAMK2B

calcium-calmodulin-dependent protein kinase type II subunit β

CAMK2D

calcium/calmodulin-dependent protein kinase type II subunit δ

CBG

corticosteroid-binding globulin

CMPK

uridine monophosphate/cytidine monophosphate kinase

COTL1

coactosin-like protein

COX5B

cytochrome c oxidase subunit-5B

CPN2

carboxypeptidase N subunit-2

CRHBP

corticotropin-releasing factor-binding protein

CTIP2

COUP-TF-interacting protein-2

DARPP32

dopamine- and cAMP-regulated neuronal phosphoprotein

DLGAP1

scaffolds disk-large associated protein-1

DNAJA1

DnaJ homolog subfamily A member-1

DRP2

dihydropyrimidinase-related protein-2

EAAT1

excitatory amino acid transporter-1

ECHS1

enoyl-CoA hydratase mitochondrial

eEFA1

eukaryotic translation elongation factor-A1

eIF4E

eukaryotic translation initiation factor-4E

eIF4H

eukaryotic translation initiation factor-4H

FAP

prolyl endopeptidase FAP

FUS

fused in sarcoma

FXR2

fragile X mental retardation syndrome-related proteins-2

GADD45G

growth arrest and DNA damage-inducible protein 45G

GAPDH

glyceraldehyde-3-phosphatase dehydrogenase

GDI

Rab GDP dissociation inhibitor α

GFAP

glial fibrillary acidic protein

GGT2P

γ-glutamyltransferase-2, pseudogene

GNAO1

guanine nucleotide-binding protein G(o) subunit α

GORASP2

Golgi reassembly-stacking protein-2

GOT1

aspartate aminotransferase cytoplasmic

GOT2

aspartate aminotransferase, mitochondrial

GPCR

G-protein coupled receptor

GRIN2B

glutamate receptors ionotropic NMDA-2B

GRP78

78kDa glucose-regulated protein

GSTP1

glutathione S-transferase P

H-FABP

fatty acid-binding protein, heart

HNRNPA1

heterogenous nuclear ribonucleoprotein-A1

HSC70-ps1

heat-shock cognate protein 70 pseudogene-1

HSC70

heat-shock cognate protein-70

HSC71

heat shock cognate 70kDa protein

HSP7C

heat shock cognate 71kDa protein

HSPA12A

heat shock 70 kDa protein-12A

HSPA5

endoplasmic reticulum chaperone BiP

IDH2

isocitrate dehydrogenase [NADP], mitochondrial

IGF2

insulin-like growth factor-2

IGFBP3

insulin-like growth factor-binding protein-3

IGHA1

immunoglobulin heavy constant α1

IGLL1

immunoglobulin λ like polypeptide-1

ITIH3

inter-α-trypsin inhibitor heavy chain-H3

ITIH4

inter-α-trypsin inhibitor heavy chain-H4

ITPKA

inositol-triphosphate 3-kinase-A

LAMP2

lysosome associated membrane glycoprotein-2

LPA1

lysophosphatidic acid receptor-1

LYVE1

lymphatic vessel endothelial hyaluronic acid receptor-1

MAP2

microtubule-associated protein-2

MAP2K1

dual specificity mitogen-activated protein kinase kinase-1

MAP6

microtubule associated protein-6

MAPRE1

microtubule-associated protein RP/EB family member-1

MAPT

microtubule-associated protein Tau

MASP1

mannan-binding lectin serine protease-1

MBL2

mannose binding lectin

MBP

myelin basic protein

MDH1

malate dehydrogenase mitochondrial-1

MDH2

malate dehydrogenase mitochondrial-2

NDUFA1

NADH dehydrogenase [ubiquinone] 1 α subcomplex subunit-1

NDUFA9

NADH dehydrogenase [ubiquinone] 1 α subcomplex subunit-9, mitochondrial

NDUFS1

NADH-ubiquinone oxidoreductase 75 kDa subunit, mitochondrial

NDUFS5

NADH dehydrogenase [ubiquinone] iron-sulfur protein-5

NONO

non-POU domain-containing octamer-binding protein

NSF

N-ethylmaleimide sensitive factor

NT-proBNP

N-terminal pro-brain natriuretic peptide

p-SNAP25

phosphorylated synaptosomal-associated protein-25

PACN1

protein kinase C and casein kinase substrate in neurons protein-1

PARP

poly (ADP-ribose) polymerase

PBP

phosphatidylethanolamine-binding protein-1

PDE6B

rod cGMP-specific 3’,5’-cyclic phosphodiesterase subunit β

PEA15

astrocytic phosphoprotein PEA15

PFKM

ATP-dependent 6-phosphofructokinase, muscle type

PKA

protein kinase-A

PKC

protein kinase-C

PLTP

phospholipid transfer protein

PON1

serum paraoxonase/lactonase-1

PON3

serum paraoxonase/lactonase-3

PPARα

peroxisome proliferator-activated receptor α

PPIA

peptidyl-prolyl cis-trans isomerase-A

PPP1

serine-threonine protein phosphatase-1

PPP2A

serine-threonine protein phosphatase-2A

PRP16

pre-mRNA splicing factor ATP-dependent RNA helicase-16

PSD95

postsynaptic density protein-95

PTAFR

platelet-activating factor receptor

RBM14

RNA-binding protein-14

SAHH2

S-adenosylhomocysteine hydrolase-like protein-1

SERPINA3K

serine protease inhibitor-3K

SERPINE2

serpin peptidase inhibitor-E

SFPQ

splicing factor proline and glutamine rich

SGTA

Small glutamine-rich tetratricopeptide repeat-containing protein-α

SIAH2

E3 ubiquitin-protein ligase SIAH2

SIRT2

NAD-dependent protein deacetylase sirtuin-2

SLC25A11

mitochondrial 2-oxoglutarate/malate carrier protein

SLC6A11

solute carrier family 6A11

SNARE

soluble N-ethylmaleimide-sensitive factor attachment protein receptors

SPNA-2

spectrin αII chain

SRSF1

serine/arginine-rich splicing factor-1

STAT6

signal transducer and activator of transcription-6

STIP1

stress-induced phosphoprotein-1

SynGAP1

Ras/Rap GTPase-activating protein-1

TDP43

TAR DNA-binding protein-43

TFII-I

transcription factor II-I

TIF1B

transcription intermediary factor-1B

UCH-L1

ubiquitin carboxyl-terminal hydrolase-1

UQCRB

cytochrome b-c1 complex subunit-7

UQCRQ

cytochrome B-C1 complex subunit-8

VAMP2

vesicle-associated membrane protein-2

VCAM1

vascular cell adhesion protein-1

VCP

transitional endoplasmic reticulum ATPase

VDAC1

voltage-dependent anion-selective channel protein-1

vWF

von Willebrand factor