Proteomic profiling of proteins in the dorsal horn of the spinal cord in dairy cows with chronic lameness

Chronic lameness affects bovine welfare and has a negative economic impact in dairy industry. Moreover, due to the translational gap between traditional pain models and new drugs development for treating chronic pain states, naturally occurring painful diseases could be a potential translational tool for chronic pain research. We therefore employed liquid chromatography tandem mass spectrometry (LC-MS/MS) to stablish the proteomic profile of the spinal cord samples from lumbar segments (L2-L4) of chronic lame dairy cows. Data were validated and quantified through software tool (Scaffold® v 4.0) using output data from two search engines (SEQUEST® and X-Tandem®). Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) analysis was performed to detect proteins interactions. LC-MS/MS identified a total amount of 177 proteins; of which 129 proteins were able to be quantified. Lame cows showed a strong upregulation of interacting proteins with chaperone and stress functions such as Hsp70 (p < 0.006), Hsc70 (p < 0.0079), Hsp90 (p < 0.015), STIP (p > 0.0018) and Grp78 (p <0.0068), and interacting proteins associated to glycolytic pathway such as; γ-enolase (p < 0.0095), α-enolase (p < 0.013) and hexokinase-1 (p < 0.028). It was not possible to establish a clear network of interaction in several upregulated proteins in lame cows. Non-interacting proteins were mainly associated to redox process and cytoskeletal organization. The most relevant down regulated protein in lame cows was myelin basic protein (MBP) (p < 0.02). Chronic inflammatory lameness in cows is associated to increased expression of stress proteins with chaperone, metabolism, redox and structural functions. A state of endoplasmic reticulum stress and unfolded protein response (UPR) might explain the changes in protein expression in lame cows; however, further studies need to be performed in order to confirm these findings.


Introduction
Chronic pain represents a dysfunction of the nervous system [1,2], and similar to humans, this type of pain impact negatively the quality of life of affected animals [3]. Naturally a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

Animals
Twelve dairy cows were selected from a commercial dairy farm (Agricola Los Ríos, single livestock role 10.5.07.0760). All selected animals were Kiwi cross and 2 years of age or older, originating in different herds within the same farm with similar breeding, feeding and other routine practices. Spinal cord samples from 5 lame dairy cows (Lame group; n = 5) with a hind limb lameness history of at least 5 months were obtained after euthanasia via intravenous general anesthesia and intrathecal lidocaine injection at the atlanto-occipital foramen as previously reported. Similarly, spinal cord samples from 7 non-lame animals (Control Group; n = 7) without apparent lameness and without history of previous lameness episodes were selected from a commercial slaughterhouse (Frigorifico Balmaceda SPA, single livestock role 10.512.0882) after euthanasia by mechanical stunning and exsanguination according to national regulations. Researchers were not involved in the decision for euthanasia or slaughter.

Lameness assessment
Given that bovine lameness is a multifactorial condition, cows were selected considering the most prevalent causes of lameness in Southern Chile, including white line disease, sole hemorrhage, sole ulcer and digital dermatitis [31]. Lameness was confirmed and classified according to the mobility score previously described by Reader et al. (2011) [32]. Briefly, cows were classified into two groups (Lame or Control), and lameness was scored as follow: MS 0 not lame; MS 1 imperfect mobility/uneven; MS 2 impaired mobility/mildly lame; and MS 3 severely impaired mobility/very lame. Lame group consisted only of cows with a MS of 3. Exclusion criteria for both groups included the presence of visible acute wounds, visible neurological gait alteration (central or peripheral ataxia) and acute or chronic mastitis.

Spinal cord sampling and protein extraction
Lumbar spinal cord sections (L2-L4) were aseptically obtained post-mortem after removal of the dorsal aspect of lumbar vertebrae. A 20 cm segment was obtained from each animal. Dura mater and arachnoids meninges were gentle dissected and after carefully washing the tissue with cold phosphate buffered saline (PBS) samples were snap frozen in liquid nitrogen and transported to laboratory for further processing. Several segments of approximately 300 mg of the ipsilateral dorsal horn were stored in a mixture of 1 mL of PBS and 10 uL of protease inhibitor. For protein extraction, samples were sonicated three times for 30 seconds each and then centrifuged at 20,000 x g for 20 min at 4˚C. The supernatant was collected and stored at -80˚C until further analysis. Total protein quantification was performed using the BCA Assay (Pierce TM Thermo Scientific, Rochford USA) according to the manufacturer's instructions. Protein extraction was evaluated using a 5-12% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) stained with Coomassie blue.

Sample preparation for proteomics analysis
100 μg of protein were lyophilized for 120 minutes and then re-solubilized in 6M guanidine hydrochloride and 25 mM NH 4 HCO 3 at pH 7.5 during 60 minutes. Subsequently, proteins were reduced at room temperature using 2 mM dithiothreitol and alkylated with 10 mM iodoacetamide at room temperature for 60 minutes. These reactions were then diluted seven times with 25 mM NH 4 HCO 3 at pH 7.5. Modified trypsin 1:50 (Promega, Madison, WI, USA) was added, and the reaction was incubated at 37˚C for 16 hours. Trypsin digestion was stopped adding 5% formic acid, pH 3.0.

Protein identification by tandem mass spectrometry (LC-MS/MS)
Protein samples were concentrated using a CentriVap concentrator (Labconco, Kansas City, MO, USA) to a final volume of 20 μL and loaded into a 350 μm internal diameter (ID) fused silica peptide trap column with 3 cm of C18 reverse-phase desalting resin. Immediately after washing with 0.1% formic acid for 30 min at 0.5 μL/min, the efflux of the peptide trap column was directed to a 10 cm resolving reversed-phase column (100 Å, 5 μm Magic C18 particles, Michrom Bioresources), which was mounted on the electrospray stage of a VELOSPRO mass spectrometer (LTQ, Thermo Finnigan LLC) by a 0-90% acetonitrile gradient for 120 min at a flow rate of 600 nL/min. An electrospray voltage of 2.2 kV with the ion transfer temperature set to 270ºC was used. The mass spectrometer was controlled by the Xcalibur software, which continuously performed mass-scan analysis of the six most intense ions during MS/MS scans of the ion traps. For this, one repeat scan of the same ion was dynamically excluded, using a 30 sec repeat duration and 90 s exclusion duration. Normalized collision energy for the MS/MS was set to 35%.
Database search algorithm and protein identification criteria. SEQUEST (Thermo Fisher Scientific, San Jose, CA, USA) and X!Tandem software (www.thegpm.org) were used to analyse all tandem mass spectra MS/MS samples. SEQUEST was set up to search Bostaur-os_NCBI_250722016.fasta (60.090 entries) with the inclusion of trypsin. Similarly, X!Tandem was set up to search a reverse concatenated subset of the uniprot-bos+taurus database (200 entries) also assuming trypsin digestion. Both databases were searched with a fragment ion mass tolerance of 0.80 Da and a parent ion tolerance of 2.5 Da. Carbamidomethyl of cysteine was specified in SEQUEST and X!Tandem as a fixed modification. Deamidated of asparagine and glutamine and oxidation of methionine were specified in SEQUEST as variable modification. Glu!pyro-Glu of the n-terminus, ammonia-loss of the n-terminus, gln!pyro-Glu of the n-terminus, asparagine and glutamine deamidation, methionine oxidation were specified in X!Tandem as variable modifications.
In order to validate MS/MS peptide and protein identification, data were loaded into Scaffold v.4.8.6 (Proteome Software Inc., Portland, OR). Peptide and Protein probabilities were assigned using the peptide/protein prophet algorithm [33]. Peptide and protein criteria of identification included a 95% confidence in the peptide/protein prophet algorithm, with a minimum of 4 identified peptideshttps://paperpile.com/c/QlNkzH/LDZX [34]https:// paperpile.com/c/QlNkzH/oLuB. Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis were grouped to satisfy the principles of parsimony. Proteins sharing significant peptide evidence were grouped into clusters. Proteins were there annotated with gene ontology (GO) terms from goa uniprot_all.gaf (downloaded Jan 12, 2017) [35]. Unique peptide counts are included as supplemental information (Supp 1).
Additionally, a search tool for the retrieval of interacting genes/proteins (STRING) neighborhood analysis was performed on all identified proteins by searching the STRING database to detect possible interactions (https://string-db.org). Protein names were obtained by matching the gen-info identifier number (GI) of the proteins to the UniProtKB database (https:// www.uniprot.org). STRING was set to identify the evidence of the type of interaction, selecting database and experiments as active interaction sources and a minimum required interaction score of 0.900 [36].s

Western blot analysis
Western blot analysis was performed in order to validate some proteins of interest detected in proteomics. Protein extraction was performed using the same method described previously for LC-MS/MS. Proteins were quantified by BSA Assay (Pierce TM Thermo Scientific Rockford USA) and twenty microgram of each sample was separated by SDS-PAGE and transferred to a 0.45 μm nitrocellulose membrane (Amersham™ Potan™ GE Healthcare Germany). Membranes were blocking during 1 hour with TBS (10 mM Tris-HCl pH 7,6) and 5% BSA at room temperature. Membranes were incubated for 16 h at 4˚C with the following antibodies: mouse monoclonal anti gamma-enolase-enolase (1:500 w/v, Santa Cruz Biotechnology), mouse monoclonal anti-Hsp70 (1:500 w/v, Santa Cruz Biotechnology) and mouse monoclonal anti-β-actin (1:500 w/v, Santa Cruz Biotechnology) as loading control. Following primary antibody incubation, membranes were washed three times (10 minutes each) with TBS and 0.1% Tween-20, and incubated during 60 minutes at room temperature with anti-mouse IgG HRP-secondary antibody (1:2000 v/w, Santa Cruz Biotechnology). Following secondary antibody reaction membranes were washed three times for 10 minutes each. The intensity of the bans were visualized with enhanced chemiluminescence (Pierce ECL Thermo Scientific Rockford USA) and captured by Odyssey Imaging System (LI-COR Bioscience, Lincoln, NE). For quantification of blot intensities Image Studio Lite ver 5.2.5 software (LI-COR Bioscience, Lincoln, NE) was used.

Statistical analysis
Label free total spectra counting and total abundance of proteins for lame and control cows were tested for normality using the Kolmogorov-Smirnov test and then compared between groups using the t-test. Band intensities, expressed as mean ±SEM, were tested for normality using the Kolmogorov-Smirnov test and then compared between groups using the t-test. Differences were considered significant when p � 0.05. All statistical tests were performed on Graphpad Prism 7.0.

Results
Tandem mass spectrometry analysis was able to identify a total amount of 177 proteins in the dorsal horn of the spinal cord. Proteins with significant peptide evidence were grouped into 47 clusters. From the totality of identified proteins, Scaffold software was able to quantify 129 of them. Quantitative analysis showed that 10 proteins were significantly downregulated, and 27 proteins were significantly upregulated in cows with chronic lameness compared to controls (Tables 1 and 2). Moreover, 26 proteins were only detected in the dorsal horn of the spinal cord of lame cows (Table 2).
The identified and quantified proteins were subjected to STRING interaction analysis. Out of 129 proteins, 121 were successfully recognized in the STRING database and represented as a network of proteins connected with evidence-based edges. The resulting network shows four evident clusters in which one protein is connected to at least three other proteins with highest confidence bond (Fig 2). One cluster consisted of 10 proteins identified as chaperones and cochaperones from which 7 were only detected in lame cows (Fig 2A). A second cluster consisted of 16 proteins associated to glycolysis, gluconeogenesis and the pentose phosphate pathway (Fig 2B), from which 5 of them were upregulated in lame cows (Table 3). Also, a third cluster ( Fig 2C) consisting of mitochondrial proteins and a fourth cluster (Fig 2D), consisting of tubulin isoforms showed no differences between lame and non-lame cows. A marked interaction between clusters A and D (Fig 2) could be observed. The Kyoto Encyclopedia of Genes and Genome (KEGG) pathways most representative in the interacting network are depicted in Table 3. The most notorious finding is that antigen processing and presentation pathway (KEGG pathway ID 0416) was only integrated by proteins strongly upregulated in lame cows.
Validation of LC-MS/MS results was performed by western blot analysis of one protein from the chaperone cluster (Hsp70) and one protein from the glycolytic enzymes cluster (gamma-enolase) (Fig 3).

Discussion
In this study we describe the differential expression of proteins in the dorsal horn of the spinal cord of cows with chronic painful inflammatory lameness using a proteomic analysis. Peptide spectra were identified using SEQUEST search engine and validated through X!Tandem using different modification parameter in order avoid identification redundancy. The totality of the lame dairy cows included in our study had a history of lameness of more than 5 months with notorious and evident hyperalgesia and allodynia of the affected limb during the mobility score evaluation. Lame cows had a higher number of expressed proteins in the dorsal horn of the spinal cord compared to sound cows. Different proteomic profile studies in dorsal root ganglia and spinal cord in various animal models of pain have previously reported an increased number of upregulated proteins compared to downregulated proteins [37].
We describe an increased amount of stress-associated proteins in the spinal cord of lame cows. A 20-time fold change in the constitutive Hsc70 (HSPA8) protein was observed in lame cows, with the inducible forms of Hsp70 (HSPA1L) and Hsp90 (HSPA90AA1) only detected in lame cows. Similar to our findings, several proteomic profiles in chronic neuropathic pain in rodents have reported increasing levels of heat shock proteins (HSP´s) after injury in both the peripheral and central nervous system [37,38,39]. Additionally, upregulation of HSP's has been reported after experimental lower lumbar and sacral terminal nerve compression in dogs [40]. HSP's are a family of intracellular chaperones that bind to different proteins facilitating their folding into their native and active conformation [41]. In response to cellular stress, upregulation of HSP's help to protect cells from abnormal protein aggregation, thus preventing cell death [42,43]. However, newer roles of HSP´s have been described recently, including antigen presentation, immune cell activation and inflammation [44]. Furthermore, neurons, glial and Spinal proteomic profile of chronic lame dairy cows   Edges represents evidence of the type of interaction, selecting database, experiments coexpression and co-occurrence as active interaction sources with a high confident of interaction score (0.900). Circles distinguish four groups of interaction. A: consisted in 10 proteins identified as chaperones and co-chaperones in which 7 were only detected in lame cows and 1 was strongly upregulated in lame cow. B: conformed by 16 proteins involved in glycolysis, gluconeogenesis and pentose phosphate pathway. C: composed mainly by mitochondrial proteins involved in the tricyclic acid pathway with a relative even expression in both groups. D: interaction generated by tubulin isoforms with no difference in the level of expression among groups but interacting with chaperone network. Light blue edges represent known interactions obtained from curated data base, violet edges represent known interactions experimentally determined, and black edges represent protein association by co-expression.    Hsp90 in DRG in a rat arthritic model, describing that the pharmacological inhibition of Hsp90 induced analgesia and a reduction in astrocyte activation. Furthermore, increased levels of Hsp90 have been described in different chronic inflammatory diseases [58] and Hsp90 has been associated with the development of neurodegenerative diseases due to its role in stabilizing the transcriptional factor NF-κB [59]. The increased number of chaperone proteins here described could be associated with endoplasmic reticulum (ER) stress inducing an unfolding protein response (UPR) [43,60]. ER stress promotes cell protection against different insults by altering the transcriptome and proteome. However, prolonged ER stress may disrupt the protective mechanisms of UPR, leading to the activation of inflammatory and apoptotic processes [43]. In our study, the ER chaperone Grp78 (also known as BiP or HSPA5) was strongly upregulated in lame cows. Grp78 has been described as a marker of ER stress and UPR [61]. UPR is associated with an increased demand of protein folding and secretion in the ER, which could explain the elevated number of chaperone proteins found in lame cows. UPR has been reported in several neuroinflammatory conditions in which protein aggregates are found [43]. Recently, Zhou et al. (2017) [62] reported an increased expression of Grp78 in the dorsal horn of the spinal cord in a rat model of inflammatory pain. Increased levels of Grp78 have also been reported in an orofacial pain model in rats [63] and in the spinal cord after L5-spare nerve ligation (SNL) in a rat model of neuropathic pain [64]. Interestingly, increased expression of Grp78 has been shown to participate in the activation of astrocytes and microglia [65,66]. Moreover, chronic morphine administration causes an upregulation of Grp78 and UPR leading to morphine tolerance in rats [67]. This finding may explain in part the mechanisms involved in the lack of analgesic effect of opioids during chronic pain states or might be involved in the downregulation of the endorphin antinociceptive descending pathway described during chronic pain.
Another cluster of interacting proteins detected in the STRING analysis corresponded to metabolic enzymes. Additionally, GO and KEGG pathway analysis showed a marked number of proteins involved in the glycolytic pathway, tricarboxylic acid pathway and the pentose phosphate pathway. Similar to our findings, changes in the expression of metabolism-associated proteins are frequently reported in pain studies using a proteomic approach [37, 38, 39]. These findings could be explained by the increasing metabolic demand of the neural tissue following nerve damage, since protein synthesis is an energy demanding process [68]. Enolase is a glycolytic and multifunctional enzyme that is frequently detected in proteomic studies [69]. Here, enolase 1 (α-enolase; ENO1), enolase 2 (γ-enolase or neuronal specific enolase; ENO2) and enolase 3 (β-enolase; ENO3) were strongly upregulated in lame cows. Non-glycolytic functions of metabolic enzymes have gained attention, in particular enolase, which can act as plasminogen receptor during pathological states [70]. Inflammatory signals promote enolase translocation to the cell membrane [71] enhancing plasminogen activation [72] and promoting extracellular matrix degradation, metalloproteinase activation, macrophage migration and cytokines synthesis [71,73]. Cell surface enolase has been found in neurons, activated microglia [74], astrocytes [73] and different immune cells [71]. The role of enolase in chronic pain was confirmed by Polcyn et al. (2017) [75], after selectively blocking its non-glycolytic functions, thus reducing spinal glial activation and cytokine synthesis in a rat model of spinal cord injury.
Antioxidant proteins SOD1, GST, PRDX1 and PRDX4 were significantly upregulated in lame cows. Different modifications in the expression of antioxidant proteins have been reported in pain studies using proteomics analysis [38,39]. Reactive oxygen species (ROS) increase in the spinal cord during chronic pain [76] participating in the phosphorylation of the NMDA receptor [77], in the activation of TRPA1 and TRPV1 channels [78] and decreasing GABA release by inhibitory interneurons [79]. Furthermore, ROS signaling is an important regulator of ER stress and UPR [80,81] and acts as a potent stimulus for chaperone proteins upregulation [82], which is consistent with the findings previously described.
Peroxiredoxin 1 (Prdx1) and Peroxiredoxin 4 (Prdx4) are active mediators of the ROS signaling pathway [83]. Prdx1 has previously been reported to increase during chronic pain [39]. Additionally, Prdx1 can be secreted by non-classical pathway, activating the immune response [84]. In contrast, Prdx4 was only detected in the spinal cord of lame cows suggesting a high demand of antioxidant activity in the ER. Prdx4 is a peroxidase with chaperone functions predominantly localized in the ER that assist protein folding by reducing H 2 O 2 , thus preventing oxidative stress in the ER [85]. Increased levels of Prdx have been associated with chronic inflammatory conditions and neurodegeneration [86]. We believe that Prdx proteins may play an important role in the development of chronic pain of lame cows.
Fascin is another protein that was strongly upregulated in the dorsal horn of lame cows. Fascin is an actin-bundling protein that regulates cell motion [87]. Fascin upregulation in the CNS promotes microglial activation; cell migration and proinflammatory cytokine release [88]. Similarly, dihydropyrimidinase-like proteins (Dpysl) participate in synaptic formation during CNS development [89] interacting with actin and microtubules [90]. Two proteins of this family were differentially expressed in lame cows. Dpysl2 was downregulated and Dpysl3 was strongly upregulated. Dpysl2 are frequently downregulated in proteomics studies after nerve damage in order to control axonal guidance [26]. In contrast, microglial activation leads to an increase in the expression of Dpysl3 following stimulation by LPS [91]. Additionally, microtubule-associated protein 1A (MAP1A) and 1B (MAP1B) were only present in lame cows. Both proteins are expressed along the axon and dendritic processes of neurons, both in the CNS and the peripheral nervous system (PNS), where they bind to microtubules and microfilaments [92]. MAP1B is only expressed during embryogenesis and its role in nervous system development has been described [93,94]. A potential role of MAP1B in the mature CNS has not been defined, but its upregulation could be associated with the presence of protein aggregates in neurodegenerative disease, proteosomal degradation and autophagy [94,95].
A limitation of our study is that several upregulated proteins in lame cows were not integrated in the enrichment pathway analysis. This could be explained by the incomplete Bos taurus annotation database due to the reduced scientific information of the species and to the highest confidence score selected in order to reduce false positive rate. This must be taken into account when bioinformatics analyses are performed in less common species. Also, the small sample size in the lame group must be taken into account before extrapolating these results. Furthermore, the proteomic results here presented must be confirmed using additional functional molecular analysis.

Conclusions
The results here presented demonstrate that persistent pain originated by chronic inflammatory lameness in dairy cows is partly mediated by ER stress. Evident changes in chaperones, metabolic and redox proteins that are frequently upregulated under cellular stress are described. Moreover, the proteome of the dorsal horn from chronically lame cows showed increased expression of several proteins with non-canonical functions. This non-canonical function might be triggered by stress signals originated from the persistent painful stimulus. Reactive oxygen species, ER stress and UPR are known to play an important role in the maintenance of chronic pain states. Further molecular analysis is necessary in order to confirm the findings here described.