Integrated Proteome and Phosphoproteome Analyses Reveal Early- and Late-Stage Protein Networks of Traumatic Brain Injury

Traumatic brain injury (TBI) is a major public health concern all around the world. Accumulating evidence suggests that pathological processes after brain injury continuously evolve. Here, we identified the differentially expressed proteins (DEPs) and differentially expressed phosphoproteins (DEPPs) in the early and late stages of TBI in mice using TMT labeling, enrichment of Phos affinity followed, and high-resolution LC-MS/MS analysis. Subsequently, integrative analyses, including functional enrichment-based clustering analysis, motif analysis, cross-talk pathway/process enrichment analysis, and protein-protein interaction enrichment analysis were performed to further identify the different and similar pathophysiologic mechanisms in the early and late stage. Our work reveals a map of early and late-stage protein networks in TBI, which shed light on useful biomarkers and the underlying mechanisms in TBI and its sequelae.


Introduction
Traumatic brain injury (TBI) is a major public health concern all around the world. In particular, China has more patients with TBI than most other countries in the world, with a mortality of approximately 13 cases per 100,000 people (Jiang et al. 2019). TBI may lead to various clinical states in the early and late stages, such as acute seizures and epilepsy in the early stage, neuroendocrine dysfunction, depression, post-traumatic stress disorder, chronic traumatic encephalopathy, and so on in the late stage (Morganti-Kossmann et al. 2019;Graham et al. 2020;Stein et al. 2019;Wilson et al. 2017). Accumulating evidence suggests that pathological processes after brain injury continuously evolve (Wilson et al. 2017;Green 2015;Palacios et al. 2020). Therefore, it's necessary to further illustrate the development of TBI, to compare the different and similar pathophysiologic mechanisms in the early and late stages.
Brain damage in TBIs is caused by rotational or linear acceleration forces, or by blunt trauma with impact deceleration (Blennow et al. 2016), and sometimes even by penetrating head injury (Luo and Fei 2015). No matter what kind of damage, proteins play a critical role in the development of pathophysiologic progress after TBI because proteins are the direct executors of biological functions. For example, postsynaptic scaffold protein Preso promotes signaling which goes through from NMDAR to NO and thus facilitates excitotoxicity after neuronal injury (Luo et al. 2019). In addition, cis-phosphorylated protein tau contributes to the pathology development of TBI and is important for the progress of Yutao Huang, Haofuzi Zhang, and Erwan Yang contributed equally to this work. sequelae after TBI (Vadhan and Speth 2021). The interactions among proteins or phosphoproteins link the different biological functions together and may be helpful for illustration of the pathophysiologic mechanisms after TBI and the therapies of TBI and its sequelae.
In this study, differentially expressed proteins (DEPs) and differentially expressed phosphoproteins (DEPPs) were identified in the early and late stages of TBI in mice using TMT labeling, enrichment of Phos affinity, and high-resolution LC-MS/MS analysis (Fig. 1A). Integrative analyses, including functional enrichment-based clustering analysis, motif analysis, cross-talk pathway/process enrichment analysis, and protein-protein interaction enrichment analysis were performed to further identify Fig. 1 Algorithm tree of the integrated analysis and validation of CCI model. A Algorithm tree of the integrated analysis. B The neurological severity score (NSS) (Shapira et al. 1988) was performed to validate the effect of the CCI model. The score consists of 10 individual clinical parameters, including tasks on motor function, alertness, and physiological behavior (Flierl et al. 2009). One point is awarded for the lack of a tested reflex or for the inability to perform the tasks, and no point for succeeding. A maximal NSS of 10 points thus indicates severe neurological dysfunction, with failure of all tasks. C The Nissl staining. The brain tissue was extracted and validated through Nissl staining. The right cortex shows obvious damage when compared with the left control the different and similar pathophysiologic mechanisms in the early and late stage, thereby contributing to a map of early and late-stage protein network in TBI (Fig. 1A). The interacting proteins may shed light on the process of pathophysiologic development and may be helpful for future therapies.

Traumatic Brain Injury Model and the Validation
Thirty-two C57BL/6 male mice (12 weeks, 25~30 g, each group: eight, total: thirty-two) that were provided by Fourth Military Medical University were fed in an air-conditioned house (27 °C) for 7 days and exposed to a 12-h light/dark cycle. The animal study was performed under the guidance of the National Institutes of Health Guide for the Care and Use of Laboratory Animals at the Fourth Military Medical University.
The cortical controlled impact (CCI) model was selected as the TBI model in vivo. The mice were placed in the stereotactic frame after anesthesia (oxygen flow: 800 mL/min, isoflurane concentration: 4%). Incise the skin in the middle of the head and carefully remove the blood and tissue on the bone to expose the skull. Identify anatomical landmarks Lambda (caudal aspect) and Bregma (frontal aspect). Draw a circle in the center of Lambda and Bregma with a 4-mm diameter and 0.5 mm away on the right side from the midline. Then, use a drill to cut along the marked circle. Note keeping the dura intact carefully. Subsequently, the exposed right cortex was hit with an actuator (3 mm in diameter) at a rate of 3 m/s and a dept of 1.5 mm which was compressed to cause damage. This would cause a degree of severe TBI (Romine 2014). Then, the vulnus was sutured, and the wound was infiltrated with lidocaine cream. At the end of the surgery, mice were returned to their cages with free access to water and food. The sham group underwent the craniotomy but didn't undergo the cortical impact. After 1 day and 7 days, the neurological severity score (NSS) (Shapira et al. 1988) was performed to validate the effect of surgery. The score consists of 10 individual clinical parameters, including tasks on motor function, alertness, and physiological behavior (Flierl et al. 2009). One point is awarded for the lack of a tested reflex or for the inability to perform the tasks, and no point for succeeding. A maximal NSS of 10 points thus indicates severe neurological dysfunction, with failure of all tasks. Further, the brain samples from the injured area (a circular region in the center of Lambda and Bregma with about 3~4-mm diameter and 0.5 mm away on the right side from the midline) were extracted and used for Nissl staining and further experiments.

Trypsin Digestion
After common protein extraction , the DTT (10 mmol/L) was used to reduce the protein liquid (37 °C, 1 h). Subsequently, the IAA (20 mmol/L) was used to alkylate the mixture (40 min, 25 °C, keep in the dark). The TEAB (100 mmol/L) was added to the protein solution to dilute the urea until 2 mol/L. After this, the trypsin (trypsin: protein = 1: 50, mass ratio) was added to the protein solution for the first digestion (8 h). The second digestion (4 h) used trypsin (trypsin: protein = 1: 100, mass ratio).

TMT Labeling and HPLC Fractionation
The Strata X C18 SPE column was used to desalt the peptide that was digested by trypsin in the 5.2 part. After vacuumdrying, the TEAB (0.5 mol/L) was used to reconstitute the peptide. Subsequently, the 6-plex TMT kit was used to process it following the manufacturer's protocol (the quantity of the TMT reagent which was enough for labeling 100 μg of the protein was defined as one unit). The mixtures were desalted and vacuumdried after the incubation of 2 h (27 °C). The Agilent 300Extend C18 column, whose internal diameter is 4.6 × 10 −3 m, length is 0.25 m, and particle size is 5 × 10 −6 m, was used to separate the specimens through high pH reverse-phase HPLC according to the manufacture's protocol.

Affinity Enrichment
For the identification of phosphoproteins only. The IMAC microspheres were added to the peptide solution following the manufacture's protocol, and then, the mixture was incubated vibrantly. After removing the supernatant, the precipitation was collected through centrifugation following the manufacture's protocol. Subsequently, the precipitation was washed by the ACN/TFA solution (50%: 6%, and then 30%:0.1%, respectively), according to the manufacture's protocol. Then, the 10% NH 3 •H 2 O elution buffer was added to the washed precipitation, and the mixture was incubated vibrantly following the manufacture's protocol. Finally, the supernatant was collected and lyophilized.

LC-MS/MS Analysis
The 0.1% FA was used to dissolve the collected peptides. The solution was loaded onto the reversed-phase precolumn. The reversed-phase analytical column was used to separate peptides. The concentration gradient was from a concentration of 7~20% B solvent (98% CAN, 0.1% FA) for 25 min, to a concentration of 20~35% for 8 min, and then up to a concentration of 80% for 3 min. Finally, the concentration was kept for 5 min. The study was performed using the EASY-nLC 1000 UPLC system, which controlled the flow rate of 300 ml/min. The Q ExactiveTM plus hybrid quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific) was used to analyze the result.
The tandem mass spectrometry (MS/MS) in Q ExactiveTM plus which was coupled online to the UPLC was used to process the peptides after the acceptance of the NSI source. The NCE was set to 30 for LC-MS/MS peptides detection, and the resolution was set to 70,000 to text the intact peptides while the resolution was set to 17,500 to detect the ion fragments in the Orbitrap. The top 20 precursor ions that exceed the threshold of 2 × 10 4 were scanned with 30-s dynamic exclusion in the MS survey system. And the procedure based on data that commutes from one MS scan to 20 LC-MS/MS scan was used to identify the top 20 precursor ions. The electrospray voltage was set to 2 kV. To keep the ion trap from overfilling, automatic gain control (AGC) was applied. The application of LC-MS/MS spectra was thresholded by the amount of 5 × 10 4 ion. The m/z scan range was set from 350 to 1800 in MS scan while the fixed first mass was set to 100 m/z.

Database Search
The Mascot search engine (v.2.3.0) for proteome and MaxQuant with an integrated Andromeda search engine (v.1.4.1.2) for phosphoproteome were used to analyze proteome and phosphoproteome respectively. The Swissport Mouse Database was used as the reference resource. For proteome, trypsin/P, which was recognized as a special cleavage enzyme, was set up to 2 missing cleavages. For phosphoproteome, the trypsin/P was set up to two missing cleavages, five modifications per peptide, and five charges. The error was set to 10 ppm and 0.02 Da for precursor ions and fragment ions, respectively. The false discovery rate (FDR) which was used for the identification of peptides, modification sites was set to 0.01. The TMT-6-plex was used for protein quantification. The peptides whose length was less than 7 amino acids were ignored.

Quality Control
The MS data validations for proteome ( Supplementary Fig. S1A, B) and phosphoproteome ( Supplementary Fig. S1C, D) were performed. The distribution of mass error is around zero, most of which are less than 0.02 Da ( Supplementary Fig. S1A, C). Furthermore, the length in proteome was distributed from 8 to 16 ( Supplementary Fig. S1B) while the length was distributed from 8 to 20 ( Supplementary Fig. S1D). The results revealed that the samples meet the criteria.

Functional Enrichment-Based Clustering Analysis
The functional enrichment analysis was based on the KEGG Pathway database, GO Biological Process database, and InterProscan database. The p-value which was less than 0.05 was deemed as significant (two-tailed Fisher's exact test). The functional enrichment categories whose p-value were less than 0.01 were further transformed through the function And then, the results were further transformed for standard Z-distribution. Finally, the Euclidean distance and average linkage clustering were used to re-cluster the transformed z-scores through one-way hierarchical clustering. The results were depicted using the "heatmap.2" function in the "gplots" package in R software.

Motif Analysis
Soft motif-x was applied for motif analysis. The sequence sites around the modify-21-mers, which come from the −10 position to the +10 position, were identified. All parameters were set as default.

Cross-talk Pathway and Process Enrichment Analysis
Cross-talk pathway and process enrichment analysis were performed based on the KEGG Pathway database, GO Biological Processes database, Reactome Gene Sets database, CORUM database, TRRUST database, and PaGen-Base database. P-value was set to 0.01, the minimum count was set as 5, and the enrichment factor was set to 1.5. The results were grouped and categorized into different clusters according to the similarities which were set to 0.3 and calculated using Kappa scores. The clusters that share a similarity were connected by edges. The map is depicted using Metascape (Zhou et al. 2019) and Cytoscape (Shannon et al. 2003).

Protein-Protein Interaction Enrichment Analysis
The protein-protein interaction enrichment analysis was performed using the BioGrid database (Stark et al. 2006). The Molecular Complex Detection (MCODE) algorithm (Bader and Hogue 2003) was used to construct and refine the protein-protein interaction network. The proteinprotein interactions are depicted using Metascape (Zhou et al. 2019) and Cytoscape (Shannon et al. 2003).

Validation of the Surgery Effect of the CCI Model
The surgery effect of the CCI model can be verified by the behavior surveillance and histology experiment. For x = −log 10 (p value ) behavior surveillance, the neurological severity score (NSS) was performed (Fig. 1B). The NSS highly correlates with the severity of brain damage and was conducted by an investigator who was blinded to the experimental groups (Shapira et al. 1988). The score consists of 10 individual clinical parameters, including tasks on motor function, alertness, and physiological behavior (Flierl et al. 2009). One point is awarded for the lack of a tested reflex or for the inability to perform the tasks, and no point for succeeding. A maximal NSS of 10 points thus indicates severe neurological dysfunction, with failure of all tasks. The result shows that the CCI group has a higher score than the sham group no matter in 1 day or 7 days after the CCI (Fig. 1B). In particular, the CCI group in 1 day has a much higher score than the CCI group in 7 days, which indicated a gradual recovering after the surgery. For histology validation, the Nissl staining was performed (Fig. 1C) where the right region that underwent the CCI exhibits an obvious deficit when compared with the left control.

Identification of DEPs and DEPPs in Early and Late Stages of TBI
In proteome profiling, 4586 protein groups were identified. The cutoff of fold-change was set as more than 1.2 or less than 0.83. Totally, 3423 proteins were quantified, where 123 proteins are high-expressed, and 49 proteins are lowexpressed in group 1d when compared to the sham group (1d/C), and 109 proteins are high-expressed and 9 proteins are low-expressed in group 7d when compared to the sham group (7d/C), and 129 proteins are high-expressed and 88 proteins are low-expressed in group 7d when compared to the group 1d (7d/1d) ( Fig. 2A).
The TMT labeling, enrichment of Phos affinity, and high-resolution LC-MS/MS analysis were applied to quantitative phosphoproteomics analysis. Among the 2229 protein groups, 5961 phosphorylation sites were recognized, where 4095 sites in 1656 proteins were quantified. The cutoff of fold-change was set as more than 1.2 or less than 0.83. Among the quantified proteins, 315 proteins are high-expressed and 453 proteins are low-expressed in group 7d/C, 360 proteins are high-expressed and 738 proteins are low-expressed in group 1d/C, and 682 proteins are high-expressed and 508 proteins are low-expressed in group 7d/1d (Fig. 2B).

Functional Enrichment-Based Clustering Analyses of DEPs and DEPPs at Early and Late Stage in TBI
The functional enrichment analyses based on DEPs (Fig. 2C) or DEPPs (Fig. 2D), including biological process, molecular function, cellular component, and KEGG pathway analyses, were clustered to compare the difference in group 1d/C, 7d/1d, and 7d/C. Based on DEPs, the different biological processes in group 1d/C are multi-organism process, regulation of hydrolase activity, and acute-phase response, etc., while the different biological processes in group 7d/C are neuron projection regeneration, regulation of protein modification, etc., and the different biological processes in group 7d/1d are regulation of cell growth and cholesterol transport, etc., suggesting a process from acute-phase response to neuron repairing after brain trauma. The different molecular functions in group 1d/C are serine-type endopeptidase inhibitor activity, phospholipase inhibitor activity, etc., while the different molecular functions in group 7d/C are phosphatidylcholine-sterol O-acyltransferase activator activity, etc., suggesting a modification of phosphorylation after brain trauma. The different cellular components in group 1d/C are extracellular vesicles and exosomes, etc., while the different cellular components in group 7d/C are fibrinogen and chylomicron, etc., suggesting a transformation from early cell-cell communications to late coagulation.
Based on DEPPs, the different biological processes in group 1d/C are the establishment of protein localization to the plasma membrane, cell migration, and actin filament bundle assembly, etc., while the different biological processes in group 7d/C are neuron projection, signaling, etc., and the different biological processes in group 7d/1d are negative regulation of MAPK activity, etc., suggesting an early cell-cell communication that mediates signalings to result in neuron repairment. The different cellular components in group 1d/C are leading edge membrane, etc., while the different cellular components in group 7d/C are adhering junction, synapse, etc. The different KEGG pathways in group 1d/C are VEGF signaling, oxytocin signaling, Rap1 signaling, etc., while the different KEGG pathways in group 7d/C are calcium signaling, NOD-like receptor signaling, retrograde endocannabinoid signaling, etc., and the different KEGG pathway in group 7d/1d is insulin resistance.

Motif Analysis of the Phosphosites
We identified 34 conserved motifs based on serine (S) phosphosites, 8 conserved motifs based on threonine (T) phosphosites, and 3 conserved motifs based on tyrosine (Y) phosphosites (Fig. 3A) Further, the heatmaps were depicted to illustrate the enrichment or depletion of specific amino acids around the phosphosites of the serine (S) (Fig. 3B), threonine (T) (Fig. 3C), and tyrosine (Y) (Fig. 3D). The amino acids asparticacid (D), glutamicacid (E), lysine (K), and arginine (R) tended to be present in the proximity of serine phosphosites. The amino acids asparticacid (D), glutamicacid (E), proline (P), serine (S), and arginine (R) tended to be present in the proximity of threonine phosphosites. The amino acids asparticacid (D), glutamicacid (E), lysine (K), valine (V), and arginine (R) tended to be present in the proximity of tyrosine phosphosites. Besides, arginine (R) was increasingly presented at the sites around the serine phosphosites and threonine phosphosites but was strikingly depleted at +1 and +2 positions (Fig. 3B, C). Interestingly, arginine (R) was inclined to be presented at the sites surrounding the tyrosine phosphosites while being depleted at −1 and +1 positions (Fig. 3D). Besides, lysine (K) was greatly presented at the sites surrounding the tyrosine phosphosites but was strikingly depleted at +1 and +3 positions (Fig. 3D). Considering that we identified the different molecular functions in group 1d/C are serine-type endopeptidase inhibitor activity, phospholipase inhibitor activity, while the different molecular functions in group 7d/C are phosphatidylcholinesterol O-acyltransferase activator activity, these special amino acids near the phosphosites may reflect the preferable enzymes that catalyze phosphorylation after the brain trauma.

Common Cross-talk Pathway/Process and Protein-protein Interacting Analyses of DEPs and DEPPs at Early and Late Stage in TBI
The common cross-talk pathway/process enrichment analysis of DEPPs (Fig. 4A) and DEPs (Fig. 4B) and protein-protein interaction enrichment analysis of DEPPs (Fig. 5A) and DEPs (Fig. 5B) were performed to further identify the connections and similarities among different functional clusters.
For cross-talk pathway and process enrichment analysis of DEPPs (Fig. 4A), the identified pathways and processes were clustered according to their similarity functions. The top 20 clusters were further clustered into four separate super-clusters in group 1d/C, five separate super-clusters in group 7d/C, and three separate super-clusters in group 7d/1d. For cross-talk pathway and process enrichment analysis of DEPs (Fig. 4B), the top 20 clusters further clustered into a super-cluster in group 1d/C and three separate super-clusters in group 7d/C and group 7d/1d.
For protein-protein interaction enrichment analysis of DEPPs (Fig. 5A), the interactive proteins clustered into three main clusters in group 1d/C, where PCE/CE pathway, WNT signaling, chemical synapse transmission, small GTPasemediated signaling, and RAS signaling are important. In group 7d/C, the main cluster is the synapse and post-synapse-related proteins, suggesting the importance of chemical synapse transmission. In addition, in group 7d/C, the cluster regulation of NMDA receptors and neuron projection is also important. In group 7d/1d, the localization to synapse and post-synaptic specialization membrane is very important, suggesting consistent importance of chemical synapse and post-synapse transmission, and a transition from the signaling to neuron projection. For protein-protein interaction enrichment analysis of DEPs (Fig. 5B), in group 1d/C, the main clusters are post-translation protein phosphorylation, platelet degranulation, acute response to elevated platelet cytosolic Ca2+, and respiratory complex I biogenesis, suggesting early protein phosphorylation and acute-phase response. In group 7d/C, the main clusters are post-translational protein phosphorylation, platelet degranulation, regulation of insulin-like growth factor (IGF) transport, and uptake. In group 7d/1d, the main clusters are positive regulation of cholesterol esterification, lipoprotein particle remolding, ribosome, and SRP-dependent cotranslational protein target to the membrane, suggesting the activation of cholesterol esterification and SRP-dependent cotranslational protein target to membrane from early-stage to late-stage after TBI.

Special Cross-talk Pathway/Process and Protein-Protein Interacting Analyses of DEPs and DEPPs at Early and Late Stage in TBI
The special cross-talk pathway/process and protein-protein interacting analyses of DEPPs (Fig. 6A) and DEPs (Fig. 6B) were performed to further identify the in-depth connections behind the cross-talk pathway and process.
For DEPPs, the special cross-talk pathway/processes that are only presented in group 1d/C (involving 468 proteins) are cell part morphogenesis, membrane trafficking, synapse organization, etc., where the main protein-protein interactions are involved in UCH proteinases, mitotic prometaphase, vesicle docking, etc. The special cross-talk pathway/ processes that are only presented in group 7d/C (involving 191 proteins) are the regulation of cell morphogenesis, actin filament-based process, small GTPase-mediated signal transduction, etc. The special cross-talk pathway/processes that are both presented in group 1d/C and 7d/C (involving 505 For DEPs, the special cross-talk pathway/processes that are only presented in group 1d/C (involving 125 proteins) are regulation of the microtubule-based process, supramolecular fiber organization, neutrophil degranulation, etc., where the main protein-protein interaction is involved in respiratory chain complex I biogenesis. The special cross-talk pathway/processes that are only presented in group 7d/C (involving 71 proteins) are cell junction organization, negative regulation of cellular component organization, vacuole organization, etc. The special cross-talk pathway/processes that are both presented in group 1d/C and 7d/C (involving 47 proteins) are platelet degranulation, complement and

Neuron Projection and Recovery
The cluster neuron projection are represented both in 1d/C and 7d/C groups of DEPPs (Fig. 4A), suggesting that it's important through the early and late stage after brain trauma, which was consistent with that neuronal swelling may exert protective effects against damaging excitability in the aftermath of TBI (Sawant-Pokam et al. 2020). Small GTPase-mediated signal transduction is isolatedly represented in group 7d/C while absent in group 1d/C of DEPPs (Fig. 4A), suggesting that small GTPase-mediated signaling may play a more separate function in the late stage after brain trauma (Mulherkar and Tolias 2020). A recent study also shows that knockdown of small GTPase RAC1 could strikingly facilitate the recovery of functions in neurons (Huang et al. 2020). The cluster gliogenesis is special in group 7d/C of DEPs (Fig. 4B), suggesting a neuron repairment at the late stage after brain trauma. In fact, a recent study also reveals that IGF-1 promotes gliogenesis and improves cognitive function (Littlejohn et al. 2020). It's necessary to research whether the early-stage protein phosphorylation leads to late-stage gliogenesis in the future. For protein-protein interaction enrichment analysis of DEPPs (Fig. 5A), in group 7d/C, the cluster regulation of NMDA receptors and neuron projection is also important, which is consistent with our former work (Luo et al. 2019). More work and studies are needed to illustrate whether the earlystage PCE/CE pathway, WNT signaling, chemical synapse transmission, small GTPase-mediated signaling, and RAS signaling lead to the late-stage neuron projection. Regulation of IGF transport and uptake-related proteins was the most important interacting protein classes that were presented Fig. 4 Common cross-talk pathway/process enrichment analysis. Cross-talk pathway/process enrichment analysis of DEPPs (A) and DEPs (B); different colors represent different pathways and pro-cesses; each circle which contains at least five proteins represents a subclass of the pathway and process; cross-talk functions are connected by edges both in group 1d/C and 7d/C of DEPs (Fig. 5B), suggesting it may serve as a therapeutic target for early injury and late restoration. Recent research shows that growth hormone (GH) and IGF deficiency after TBI may inhibit the recovery of the axon and neurons, and thus, GH/IGF-I system would be benefit for the therapy of TBI (Feeney et al. 2017;Tanriverdi et al. 2015;Yuen et al. 2020).

Neurodegeneration Diseases
The cluster chemical synaptic transmission is represented both in 1d/C and 7d/C groups of DEPPs (Fig. 4A), suggesting that it may play a critical role through the early and late stage after brain trauma, which was consistent with that synaptic dysfunction plays a critical role in neurodegeneration diseases after TBI (Li et al. 2020). Modulation of chemical synaptic transmission-related proteins was identified as one of the most important interacting protein classes that were presented both in group 1d/C and 7d/C of DEPPs (Fig. 6A), suggesting signaling mediated by chemical synapse was important through early and late stage of TBI. Recent studies have also shown that TBI plays a critical role in the regulation of synapse function in the early and late stages, resulting in multiple secondary injury processes, such as excitotoxicity, inflammation, oxidative stress, and thus dementia later in life (Jamjoom et al. 2020).

Other Pathways and Functions
UCH proteinases were identified as one of the most important interacting proteins which were only presented in group 1d/C of DEPPs (Fig. 6A), suggesting they are important in the early stage of TBI and have an early diagnosis potential. A recent study reveals that serum GFAP and UCH-L1 are sensitive for the prediction of intracranial injuries (Bazarian et al. 2018). Cluster platelet degranulation represents both in group 1d/C and 7d/C of DEPs (Fig. 4B), suggesting a continuous hemorrhage. The cluster post-translational protein phosphorylation is special in group 1d/C of DEPs (Fig. 4B), suggesting protein phosphorylation at the early stage after brain trauma. Respiratory chain compex I biogenesis-related proteins (NDUFV3, NDUFS6, NDUFB7) were the most important proteins that were only represented in group 1d/C of DEPs (Fig. 6B), suggesting that energy supply is critical in the early stage in TBI. However, there's little research that focuses on these proteins in TBI. Further studies are needed to research the functions of these proteins in TBI.

Pre-clinical Prospective and Limitation
Generally speaking, four specific animal models are widely used in researching TBI: fluid percussion injury (FPI) (Dixon et al. 1987), controlled cortical impact (CCI) injury (Dixon et al. 1991), weight-drop impact acceleration injury (Marmarou et al. 1994), and blast injury (Cernak et al. 1996)]. Among these models, the CCI shows consistent outcomes in many labs all over the world because of the precise control of the damage deep, injury time, and speed of impact when compared with other TBI models (Siebold et al. 2018). An additional strength of the CCI model, when compared with models involving gravity-driven devices, is the lack of risk of a rebound injury. The histopathological severity of CCI injury is positively relative to cortical deformation and impact velocity, which allow the researchers to construct a variety of TBI models with different severity according to the requirement of the experiment (Xiong et al. 2013). Because of these reasons above, the CCI model rapidly became one of the most commonly used Fig. 6 Special cross-talk pathway/process and protein-protein interaction enrichment analysis. Special cross-talk pathway/process and protein-protein interaction enrichment analysis of DEPPs (A) and DEPs (B). In cross-talk and process enrichment analysis, different colors represent different pathways and processes; each circle that contains at least proteins represents a subclass of the pathway and process; cross-talk functions are connected by edges. In protein-protein interaction enrichment analysis, different colors represent different functions that contain at least three proteins; interacting proteins are connected by edges models of preclinical TBI. It has been considered a clinically relevant pre-clinical TBI model because of reproducing many aspects of clinical TBI. One of the most important reasons that the CCI model is clinically relevant with TBI is the similar histopathological changes in both TBI patients and CCI models. These changes include cortical contusion (Dixon et al. 1991), disruption of the blood-brain barrier (Dhillon et al. 1994), hippocampal cell loss (Adembri et al. 2014), and overall brain volume loss (Osier and Dixon 2016). Furthermore, many cascades and pathways were activated after clinical TBI. Similarly, as evidence, the CCI models were also observed these cascades and pathways, for example, apoptosis cascade (Schaible et al. 2013), inflammation pathway (Gatson et al. 2013), and oxidative stress cascade (Khan et al. 2011). Furthermore, another similarity in both clinical TBI and CCI models is the functional deficits in many domains, including overall neurological function, memory, learning, motor function, and frontal lobe function (Osier and Dixon 2016). Overall, CCI is considered a clinically relevant pre-clinical TBI model and is commonly used all over the world. However, there are still some limitations of the CCI model. Although there is substantial similarity in the physiology of Mus musculus and Homo sapiens brains, it is clear that notable differences exist between these species in terms of brain structure and function, including brain geometry, craniospinal angle, gyral complexity, and white-togray matter ratio. These structural characteristics may lead to substantially different responses to the trauma of comparable severity or type from species to species. What's more, TBI has many different types. The cerebral contusion and laceration that the CCI model cause are one type of TBI. Although cerebral contusion and laceration are the most common TBI in clinical, the CCI model cannot represent the whole TBI. In conclusion, our model shows many similar characteristics to the patients with TBI, and the model can mimic the most common TBI in clinical to a great extent. Our model sheds light on the earlystage and late-stage differences of TBI especially concerning the functional enrichment, cross-talk pathway/process, and protein-protein interactions, which help to further explore the biological mechanisms after TBI and thus a better therapy to reduce the mortality of TBI. Note that there are still some different between our model and clinical TBI where this limitation is common to many or all pre-clinical TBI models and need a further experiment to validate it.

Conclusions
We identified the DEPs, DEPPs, cross-talk pathways/processes, and interacting proteins in the early and late stages of TBI. PCE/CE pathway and β-catenin independent Wnt signaling (PSMA, SYN, etc.), chemical synapse signaling (DLG, LQSEC, etc.), and small GTPase-mediated signaling transmission pathway (GRIA, etc.), UCH proteinase signaling (PSMA, PSMB, etc.)-related proteins are the most important interacting proteins in the early stage of TBI while the NMDA signaling-related proteins (DNM, etc.) are the most important interacting proteins in the late-stage TBI. The DEPs or DEPPs in the early stage may contribute to the acute-phase response of TBI and the DEPs or DEPPs in the late stage may play a critical role in TBI sequelae.