Microdialysis Sampling from Wound Fluids Enables Quantitative Assessment of Cytokines, Proteins, and Metabolites Reveals Bone Defect-Specific Molecular Profiles

Bone healing involves a variety of different cell types and biological processes. Although certain key molecules have been identified, the molecular interactions of the healing progress are not completely understood. Moreover, a clinical routine for predicting the quality of bone healing after a fracture in an early phase is missing. This is mainly due to a lack of techniques to comprehensively screen for cytokines, growth factors and metabolites at their local site of action. Since all soluble molecules of interest are present in the fracture hematoma, its in-depth assessment could reveal potential markers for the monitoring of bone healing. Here, we describe an approach for sampling and quantification of cytokines and metabolites by using microdialysis, combined with solid phase extractions of proteins from wound fluids. By using a control group with an isolated soft tissue wound, we could reveal several bone defect-specific molecular features. In bone defect dialysates the neutrophil chemoattractants CXCL1, CXCL2 and CXCL3 were quantified with either a higher or earlier response compared to dialysate from soft tissue wound. Moreover, by analyzing downstream adaptions of the cells on protein level and focusing on early immune response, several proteins involved in the immune cell migration and activity could be identified to be specific for the bone defect group, e.g. immune modulators, proteases and their corresponding inhibitors. Additionally, the metabolite screening revealed different profiles between the bone defect group and the control group. In summary, we identified potential biomarkers to indicate imbalanced healing progress on all levels of analysis.

summary, we identified potential biomarkers to indicate imbalanced healing progress on all levels of analysis.
degradation of the proteins [12]. Additionally, a high salt content, a high complexity and a low overall concentration of proteins in the wound fluids interfere with successful comprehensive screening. Thus, proteomic analyses were limited to the quantification of only few dozen to about 150 proteins regardless of the sampling technique [17,19,20]. Recently, we established a sampling approach that allows the identification of more than 500 proteins by using a solid phase extraction of proteins adsorbed to the microdialysis catheter. Moreover, this approach can be combined with the classical continuous sampling of cytokines and of metabolites [21].
In order to identify bone-specific markers we applied microdialysis in combination with proteomics and metabolomics in two different wound scenarios in a rat bone defect (BD) and soft tissue wound (STW). Dialysates were used for analysis of selected chemokines and cytokines by ELISA to obtain time-resolved data during the first 24 hours after injury.

Animal Surgery
The study was approved by the Agency for the Inspection of Veterinary and Food Materials of the Government of Saxony (permission no. 24-9168.11-1/2013-31). All animals were housed according to the European guidelines for the care and use of laboratory animals. Forty-four male Wistar rats with an average body weight of 350 g were obtained from Janvier (Le Genest Saint Isle, France) and held in the animal care unit for at least 7 days before the experiments.
The rats were anesthetized with a solution of ketamine (100 mg/kg body weight) and xylazine (10 mg/kg body weight) and kept under anesthesia up to 12 hours (administration of ketamine/xylazine through a permanent intraperitoneal catheter every 90-120 min). After shaving and disinfection of the right hind leg a 3 cm longitudinal incision was made and the femur was surgically exposed by dissecting the thigh muscle. A five-hole plate (Stryker, Hamburg, Germany) was fixed with 4 screws (1.5 x 6 mm) on the femur. In twenty four rats a 5 mm bone defect was created using a wire saw (group: bone defect, BD). In twenty rats the bone was not injured and only a soft tissue wound was created down to the femur (control group: soft tissue wound, STW). The microdialysis catheter (CMA/20, CMA Microdialysis AB, Solna, Sweden, cut-off 100 kDa) was inserted either into the soft tissue defect near the bone or into the bone defect. The muscle was closed with absorbable sutures and the skin was closed with suture clips. The remaining animals were allowed to wake up and move freely after closure of the muscle and skin. After 12 hours these animals were anesthetized again and microdialysis catheter was inserted after opening the wound area 1 cm. The rats were covered with thin sheets to avoid hypothermia. Additionally, a mixture of 0.25 mL 0.9% NaCl solution and 0.25 mL 10% glucose was administered every 120 min through the permanent catheter to avoid hypoglycemia and dehydration. The rats were sacrificed under anesthesia at the end of the experiment by CO 2 inhalation.

Histology
At the end of the experiment the fracture hematoma was explanted, fixed with 4% formaldehyde, embedded in paraffin, and cut into 2 μm-sections for Pappenheim´s staining. For quantification mononucleated and multinucleated cells were counted in 10 randomly distributed squares per section on three subsequent slices per fracture hematoma.

ELISA
The concentration of 15 cytokines and growth factors were analyzed with commercially available ELISA kits ( Table 1). The ELISA assays were performed according the manufacturer's instructions. Mean and standard deviation were calculated when at least 3 values were above the detection limit of the relevant ELISA.

Sample Preparation and Acquisition for Proteomics
The sample preparation and acquisition was conducted as previously described [21]. Briefly, protein adsorbates (n = 4 for each wound scenario explanted after 12 h) were eluted from the catheter membrane with Laemmli buffer for 5 min at 95°C. The eluted proteins were separated by 1D-SDS-PAGE. The samples were sliced to 8 gel fractions each and an in-gel proteolytic cleavage was conducted using trypsin. 52 ng of the Proteomics Dynamic Range Standard (UPS2, Sigma Aldrich, Seelze, Germany) was spiked into each fraction. For control experiments, proteins from microdialysates were purified using Vivacon spin tubes (10 kDa MWCO, Sartorius, Göttingen, Germany). Proteins were reconstituted in Laemmli buffer, separated by a 1D-SDS-PAGE and treated as described above but without spiking-in the UPS2.
LC-MS/MS analyses were performed with a nano-HPLC system (nanoAcquity, Waters, Milford, MA, USA) online coupled to a LTQ Orbitrap XL ETD mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA). Peptide separation was achieved by RP-LC using a 90 min non-linear gradient (2-40% acetonitrile containing 0.1% formic acid). After MS full scan (positive mode, m/z 350 to 1,600, R = 60,000) MS/MS acquisitions of the six most abundant ions were automatically triggered when exceeding an intensity threshold of 3,000 counts. A dynamic exclusion was set for 120 sec limiting the list to a maximum of 500 entries. Fragmentation was conducted with the linear ion trap by collision induced fragmentation (CID, isolation window of 4 amu, normalized collision energy 35, activation time 30 ms, activation Q 0.25).

Metabolomic Profiling and Quantification by Mass Spectrometry
Reconstituted microdialysate samples from 12-24 h after injury (n = 3 for BD, n = 6 for STW, 50 μL each) were precipitated by adding 500 μL of ACN. The samples were centrifuged, evaporated and reconstituted as described by Hoeke et al. [25]. LC-MS analysis was conducted on a Thermo Fischer Orbitrap Velos system using ESI positive ionization and UHPLC separation as described previously [25]. For details see Supporting Information. In addition, full scan acquisition (m/z 100-800, resolution = 30,000) using identical MS parameters was used for quantitative analysis. Data analysis was performed by XCMS online [26] using "HPLC / Orb (136)" parameters. Detailed information on the XCMS settings were described in S1 Text.
Data obtained by XCMS analysis was confirmed by targeted LC-MS analysis using Abso-luteIDQ p150 kit (BIOCRATES Life Sciences AG, Innsbruck, Austria) to quantify 163 metabolites as described elsewhere [21]. All experiments were carried out on an Agilent 1100 series binary HPLC system (Agilent Technologies, Waldbronn, Germany) coupled with an 4000 QTRAP™ linear ion trap mass spectrometer (AB Sciex, Concord, Canada) equipped with a Tur-boIon spray source. The FIA-analysis was performed with MS running solvent at ambient temperature. Quantification was achieved by multiple reaction monitoring (MRM) detection in combination with the use of stable isotope-labeled and other internal standards. Data evaluation for quantification of metabolite concentrations was performed with the MetIQ™ software package.

Bioinformatics
Cell number was given as mean ± standard deviation (SD) and statistical significance was tested by Student t-test. The level of significance was set at p < 0.05.
ELISA data were statistically analyzed with the Kruskal-Wallis test. The Mann-Whitney test was used to detect post hoc statistical significance between selected groups. Statistical significance was set at p < 0.05.
All quantitative values for proteins gained from mass spectrometry were log10-transformed and median normalized, first between all corresponding fractions of each sample and afterwards between all samples. Proteins with quantitative values in at least three replicates of one condition and in none of the other were considered to be unique for the corresponding condition. Proteins with quantitative values in at least three replicates of both conditions were considered as quantified. To be treated as significantly higher abundant the log10-ratio in comparison of both conditions must be > 0.3 with a p-value of less than 0.05 (two-tailed, unpaired student's t-test).
Protein clustering was based on Gene ontology (GO) assignments (GO:BP_FAT, GO: CC_FAT and GO:MF_FAT) using DAVID Bioinformatics Resources 6.7 [27] and refined by manual literature research. For an enrichment analysis a list of known extracellular rat proteins (evidences gained by mass spectrometry) provided by the ExoCarta database [28] (release date 29th July 2015) was used as background list. A principle component analysis (PCA) was conducted by using R and the package ggbiplot [29,30]. Additionally, the quantitative protein data were analyzed through the use of QIAGEN's Ingenuity 1 Pathway Analysis (IPA 1 , QIAGEN Redwood City, www.qiagen.com/ingenuity). The log10-ratios and p-values of all quantified proteins were included. Unique proteins were assigned to a log10-ratio of 1 or -1, respectively, and p = 0.01.

Results and Discussion
Bone repair is a highly complex process requiring the synchronization of multiple biological pathways to restore tissue integrity. Activation of these pathways gives rise to unique wound and fracture microenvironment that consists of cells and ECM including a complex milieu of enzymes, growth factors, and cytokines that control the tissue repair response. Thus, the real therapeutic challenge in enhancing bone healing is to provide an ideal microenvironment for optimal cell migration and proliferation. Nevertheless, the whole process in its complexity is still poorly understood. Even though the significance of single proteins is determined, for a thorough understanding of the whole process it is relevant to identify and quantify the involved molecules at their site of action. It is well known that the initial inflammatory stage is essential for later consolidation [31][32][33]. Hence, the aim of this study was the identification and quantification of chemokines, cytokines and further proteins involved in the early stage of bone healing as well as metabolites that could provide further insight into the metabolic adaptions following a fracture.
Because the concentration of the cytokines in plasma does not represent the concentration in the injured tissue, quantification at the wound site is highly desirable [34,35]. Microdialysis is an ideal sampling technique as it provides the capability to monitor these proteins in real time in their surrounding milieu. However, chemokines and cytokines are small molecules present in low concentration. In addition, the BSA added to the perfusate to increase the recovery of proteins cover the signals from the chemokines and cytokines. Taken together, this leads to an underrepresented identification and quantification of those proteins in a nanoLC-MS/ MS approach [21]. Hence, ELISAs were used to quantify the cytokines in the dialysate because of their acceptable specificity, sensitivity, and the ease of performance.
Even though microdialysis is a minimally invasive technique, there is also evidence that the insertion of the catheter into the tissue can change the interstitial concentration of cytokines [36]. But since BD as well as STW are dramatic injuries the effect of implantation of the catheter in the present experiment is expected to be negligible compared to the inflammation and thus the cytokine response caused by the injuries.

Clinical Findings
Forty three rats survived the surgery without any complications. One animal died due to severe anemia during the surgery. Three animals died during the experiment without any identifiable reason. The dialysates of eleven rats with BD and those of 9 rats with STW each sampled at 0-12 h and 12-24 h could be used for ELISA analysis.

Determination of Cytokines and Growth Factors with ELISA
In total the abundance of 15 cytokines and growth factors were investigated in the dialysate from BD and STW. Under the experimental conditions six cytokines and growth factors were not detectable (BMP-2, VEGF, PDGF-BB, IL-1α, IL-1β, and TNF-α) which might be due to several reasons. First, the recovery of proteins depends reciprocally on their molecular weight [37]. Although we used a cut-off of 100 kDa the larger cytokines VEGF (MW appr. 40 kDa) or PDGF-BB (MW appr. 30 kDa) seem to be unable to cross the membrane with sufficient recovery which is in line with the observation that an acceptable recovery will only be attained if the molecular weight of the proteins is lower than approximately one-fourth of the membrane cutoff [38]. However, we used this cut-off to avoid the recovery of serum proteins and to avoid an influx of perfusion fluid into the wound area which we have seen using catheter with larger pore size. Second, the recovery is also controlled by the concentration of the protein. A higher concentration of the protein means a higher concentration gradient driving the transfer over the membrane [39]. Third, since it is well known that most cytokines are usually found in low picomolar concentrations [14] we cannot exclude that the concentration of these investigated proteins in the wound fluid was already below the detection limit. Forth, the composition of the perfusion fluid, the flow rate and the area of the membrane are further criteria influencing the recovery of proteins [40]. The optimization of the perfusion fluid composition and the flow rate we described previously [13]. But we cannot exclude that other flow rate or buffer composition could result in detectable protein concentrations. The length of the used membrane was 4 mm, because the diameter of the rat femur is about 3-4 mm. The sampling of proteins from BD could thus be limited to the area of the BD because the catheter tip does not stick out into the soft tissue. Despite these challenges and limitations, microdialysis can be used to receive both time-dependent qualitative and quantitative data from the site of injury which cannot be achieved by other discontinuous methods.
For nine markers assessed, concentration profiles could be obtained (Fig 1). IL-6, CXCL1, CXCL2, and CXCL3 showed an increase in concentration after surgery and a re-equilibration to baseline after 24 h. The concentrations of CXCL4 and CXCL7 decreased immediately after surgery, suggesting a release of the proteins from activated cells. SDF-1, TGF-β1, and IL-10 did not show any changes in concentration over the first 24 h after surgery.
The key function of IL-6 is mediation of the acute phase response. After release at the site of injury IL-6 moves to the liver through the bloodstream and induces the synthesis of acute phase proteins such as C-reactive protein, fibrinogen, serum amyloid A, haptoglobin, and α1-antichymotrypsin (SerpinA3) in hepatocytes [41]. In the present series, IL-6 concentration in the dialysate peaked earlier in the STW than in the BD. The highest concentration of IL-6 in BD was detected 12-15 h after surgery. In contrast, for STW the maximum was reached at 6-9 h. The early increase after the injury underlines the role of IL-6 in the acute phase response. With the Kruskal-Wallis test the differences in the time response of the IL-6 concentrations in STW was statistically significant different (p = 0.006). In the intervals 6-9 h and 9-12 h the absolute IL-6 concentration was significantly higher in STW than in BD (6-9 h: p = 0.047; 9-12 h: p = 0.007). A similar time response was found in UVB-irradiated skin. IL-6 increased rapidly after 8-16 h and declined to baseline after 24 h [10]. CXCL1, CXCL2, and CXCL3 are all potent chemoattractants for neutrophils and are considered to play key role in neutrophil infiltration in inflammatory diseases [42]. All three CXCLs have a similar ability to induce chemotaxis for neutrophils in vitro and in vivo [43,44]. However, the contribution of each CXCL to neutrophil infiltration in vivo seems to depend on the amount of each CXCL at the site of inflammation [45]. In addition, in rat the proportion of each CXCL at the site of inflammation depends on the type of inflammation [42]. Values are mean ± SD. IL-6, interleukin-6; CXCL2, chemokine (C-X-C motif) ligand 2; CXCL7, chemokine (C-X-C motif) ligand 7; CXCL3, chemokine (C-X-C motif) ligand 3; CXCL1, chemokine (C-X-C motif) ligand 1; CXCL4, chemokine (C-X-C motif) ligand 4; IL-10, inteleukin-10; TGF-β1, transforming growth factor-β1; SDF-1, stromal-cell derived factor-1. * p < 0.05, ** p < 0.01, *** p < 0.001.  expression has been demonstrated for platelets but also endothelial cells, neutrophils, monocytes, and macrophages [46][47][48]. In BD the highest CXCL1 concentration was detected 6-9 h after surgery. In STW the highest CXCL1 concentration was seen 3-6 h after surgery. For both defect scenarios the time response showed significant differences (BD: p = 0.007; STW: p = 0.004) resulting in p-values p < 0.01 determined by a post hoc test. In the intervals 15-18 h and 18-21 h the differences between BD and STW were statistically significant (15-18 h: p = 0.005; 18-21 h: p = 0.025). CXCL1 is constitutively expressed by endothelial cells, megakaryocytes in the bone marrow, and to a lesser extent by osteoblasts [49,50]. This might explain the higher CXCL1 concentration in BD compared to STW where the bone and thus the bone marrow were not injured.
The macrophage inflammatory protein 2 alpha (CXCL2) is not stored in platelets but expressed, among other cells, by macrophages and monocytes [51]. In contrast to the IL-6 concentration, CXCL2 in the dialysate peaked earlier in BD than in STW. The highest CXCL2 concentration in BD was detected 6-9 h after surgery while the maximum level in STW was detected 12-15 h after surgery. However, 12 h after surgery no CXCL2 was detectable in BD. The post hoc test revealed significant differences between 0-3 h and 3-6 h (p < 0.001), 6-9 h (p < 0.001) and 9-12 h (p = 0.002). The total CXCL2 concentration was higher in STW than in BD.
The CXCL3 concentration was highest 6-9 h after surgery in BD as well as in STW. In BD CXCL3 was not detectable anymore after 12 h. Nevertheless, the time response was statistically significant different in BD (p = 0.013) and STW (p < 0.001) and resulted in p-values from post hoc tests below p < 0.01.
For all three CXCL types we found a similar time response. The concentration increased immediately after injury, peaked early and then decreased after 12-15 h. For CXCL1 and CXCL2 the inflammatory response was comparable to that in a rat burn model. In the latter, serum CXCL1 increased rapidly after 3 h and then decreased slowly while CXCL2 showed a rapid increase immediately after the injury with a peak at 6 h and then a dramatic decrease between 12 h and 24 h [52] similar to the missing CXCL2 in BD after 12 h in the current experiment.
Platelet factor 4 (CXCL4) and pro-platelet basic protein (CXCL7) are abundant chemokines in the α-granules of platelets that are released during platelet activation [53]. This is in accordance with the concentration profile determined for CXCL4 and CXCL7 in the present study. In BD and STW the CXCL4 concentration decreased after surgery and reached steady-state level 9-12 h after surgery. The difference in time response was not statistically significant different (BD: p = 0.099; STW: p = 0.981).
For CXCL7 the concentration decreased from 27.3 ng/mL (SD 26.8 ng/mL) at 0-3 h to 0.8 ng/mL (SD 0.68 ng/mL) at 12-15 h and remained at low level for the rest of the sampling intervals in BD. In STW the steady state level was reached with the 6-9 h interval. However, the time response for STW is statistically significant different (p = 0.02).
A third time-dependent concentration profile was found for IL-10, TGF-ß1 and SDF-1. For these cytokines, the concentration in the samples did not change over time. IL-10 is an antiinflammatory cytokine and thus it will be upregulated at the end of the inflammatory stage. On the other hand, there is evidence that the recovery of IL-10 is low because of a larger hydrodynamic radius compared to other cytokines [40,54].
TGF-β1 is important in all wound healing stages. In the inflammatory stage it is released by platelets and facilitates the recruitment of additional inflammatory cells and MSCs [3,55]. After that TGF-β1 is being expressed by other cells over 21 days during fracture healing [3]. In STW TGF-β1 was only detectable until 12-15 h after surgery. Until then the concentration slightly increased from 9.2 pg/mL (SD 12.9 pg/mL) to 21.7 pg/mL (SD 17.6 pg/mL). For BD the TGF-β1 concentration ranged from 3 to 17 pg/mL over 24 h.
The concentration of SDF-1 in the BD samples remained constant over the whole 24 h. In STW samples the SDF-1 concentration increased after the first interval and post hoc test identified a statistically significant increase from 0-3 h to 6-9 h (p = 0.017). SDF-1 recruits MSCs during endochondral bone repair. An increase of the SDF-1 mRNA was not observed before day 2 in a mouse segmental bone graft model [56]. Thus, the constant local protein level over the first 24 h after creation of a bone defect may indicate an early steady state before systemic upregulation.
The ELISA results show that microdialysis is a suitable method to determine the concentration profiles of various proteins associated with the inflammatory stage of the healing process. In addition, we could show differences between BD and STW. This is an important fact because it is now possible to identify dysregulated proteins in delayed wound and fracture healing. Second, influencing the expression or the release of these proteins could result in an improved healing process.
The number of polymorphonuclear cells in the fracture hematoma from BD was significantly lower after 12 h compared to 24 h indicating the beginning of the inflammatory stage (Fig 2A). After 12 h we could mostly detect neutrophil granulocytes and some monocytes. In the fracture hematoma 24 h after surgery increased number of neutrophil granulocytes but also eosinophil and basophile granulocytes were seen (Fig 2B and 2C). The increase in polymorphonuclear cells corresponds to the time line of cellular migration during healing. After coagulation neutrophils are the first cells arriving at the wound site peaking after 24-48 h. They are responsible for phagocytosis and wound debridement and thus they decrease the rate of infections [57,58]. Neutrophils are important cells because they also produce and express chemokines which chemoattract further immune cells and as a result the healing process is going on. We could only analyze the fracture hematoma from BD because no hematoma was formed in a STW.

Wound Fluid Proteomics
The combination of microdialysis with proteomics provides additional information about the composition of the fracture hematoma and could be used in order to identify potential biomarkers of bone healing. In the present experiment, the total amount of protein recovered from BD or STW by microdialysis was too low for HPLC-MS/MS analysis. We identified only 28 proteins, most of them ubiquitously produced in cells (S1 Table). We therefore used the proteins adsorbed to the tip of the microdialysis catheter membrane for proteomic analysis. This allows for effective and reproducible sampling as has been shown recently [21]. By using this approach an overall number of 513 proteins could be identified in a comparative analysis of both defect scenarios. Among those 273 proteins could be reliably quantified (S2 Table).
To identify significant alterations in the protein composition of fluids from different wound scenarios, two different approaches were pursued: (i) identification of unique proteins for a certain wound scenario and (ii) a determination of significant higher abundance of proteins for a certain wound scenario. To be considered as unique, a protein must be quantified in at least three of four replicates of one wound scenario but in none of the other. To be considered as significantly higher abundant, the Log 10 -ratio of the protein abundance in both defect scenarios must exceed 0.3 (= 1-fold increase) with a p-value of p < 0.05 (student's t-test). The 19 quantified proteins of the spiked-in Proteomics Dynamic Range Standard served as control (Fig 3A). None of the spiked-in standard proteins was identified as significantly regulated, indicating a successful quantification and normalization procedure. By applying both described strategies to all 273 reliably quantified proteins, 107 of them could be assigned to a certain wound scenario ( Fig 3B). The number of specific proteins for the bone defect group (BD, 30 unique, 50 by higher abundance) was notably higher than the number for the soft tissue wound group (STW, 2 unique, 25 by higher abundance). This difference appears plausible because the BD scenario has all features of the STW but with an additional fracture of the femur.
To confirm a specific protein profile of the sampled wound fluids for a given defect scenario a principle component analysis (PCA) of all quantified proteins was conducted (Fig 4). The two defect groups could be separated by this analysis.
In summary, by (i) identifying unique proteins for both defect scenarios, (ii) showing a significant higher abundance of several proteins depending on certain wound scenarios and (iii) by the clear separation of both defect types in the PCA, defect-specific protein profiles can be considered.

Identification of Early Wound or Fracture Healing Markers
To gain deeper insight into the biological potential of the wound fluids all quantified proteins were assigned to functional clusters based on GO-term assignment and manual literature research. An enrichment analysis of all quantified proteins revealed, among others, an increased presence of proteins being involved in proteinase inhibition (enrichment score = 3.8), wounding and inflammatory response (3.29), apoptosis (2.79), tissue regeneration (2.46), blood coagulation (2.43) and lymphocyte chemotaxis and activation (2.06) (S3 Table). Moreover, a signaling pathway analysis revealed several enriched and activity-altered processes (S4 Table). A selection of higher hierarchical signaling pathways with a focus on cellular adhesion and migration ( Fig 5A) and immune response (Fig 5B) is shown. Several pathways are predicted to be in a higher activity state in the bone defect group. This could indicate an increased invasion of immune cells to the wound area after a femur fracture compared to the soft tissue wound. The acute phase response signaling is one of only two pathways with a decreased activity. This confirms the finding of an earlier IL-6 response in the soft tissue wound.
Since an impaired wound healing is believed to be based on an improper inflammatory response, we manually clustered all quantified proteins to early wound healing mechanisms including ECM proteins, immune modulators, proteases and protease inhibitors as described before [19] (S5 Table). Tables 2-5 lists the top 15 most abundant proteins including all unique or specific proteins for a certain wound scenario, assigned to those clusters.

Different Abundance of ECM Proteins in Wound Fluids
Proteins forming insoluble fibers are essential for primary wound closure. Moreover, they provide a provisional ECM for the migration and adhesion of immune cells and re-epithelization. Overall, nine proteins could be assigned to the ECM protein cluster ( Table 2). With fibrinogen alpha and gamma chain (FGA, FGG) and fibronectin (FN1) two were specific by higher abundance for the STW group. The mature fibrin and FN1 were shown to be involved in forming a provisional ECM [59,60]. The earlier formation of a provisional ECM is in line with the finding of an earlier response of the acute phase mediator IL-6 in the STW group compared to the BD group. However, alpha 2 HS-glycoprotein (AHSG) was also identified specific by higher abundance for the STW group. This protein was shown to be negatively correlated to the acute wound healing phase [61]. Galectin-3 (LGALS3), olfactomedin 4 (OLFM4), and annexin A2 (ANXA2) were specific for BD. LGALS3 has been shown to be involved in the acute inflammatory response including chemoattraction of monocytes and macrophages, as well as neutrophil activation and adhesion [62,63]. OLFM4 has been found in granules of neutrophils [64,65] and appears to be an important regulator of the innate immunity against bacterial infections [66]. On the other hand, OLFM4 was reported in higher abundance in chronic non-healing wounds [19]. ANXA2 is expressed in endothelial cells, macrophages and mononuclear cells [67]. Recent studies suggest that secreted ANXA2 can activate macrophages to secret IL-1, IL-6, and TNF-α, and can recruit and activate immune cells to the site of infection [68,69].  Table 2. Identified proteins clustered to extracellular matrix proteins. Identified proteins from proteomic analysis at 12 h after injury. Gene symbols and protein names correspond to Uniprot Knowledgebase. Quantitative values are given as Log10-transformed fold changes BD/STW. Significances are indicated as asterisks with * p-value < 0.05, ** p-value < 0.01 and *** p-value < 0.001. n = 4 for each defect scenario.

Gene symbole
Protein name Log 10 BD/STW

Different Abundance of Immune Modulators in Wound Fluids
The balance of the immune response has a critical role in influencing the outcome of wound and fracture healing. Both, an inadequate primary defense against pathogens invading the wound and an unregulated activation of the immune system overwhelming the protective capacity of the tissue could lead to a non-healing wound or fracture nonunion. 27 proteins with immunomodulatory capabilities were quantified in the wound fluids underlining the crucial role of a proper immune response at the wound or fracture site. Ten of these proteins were either unique or significantly higher abundant in the BD group (Table 3). This includes cytosolic proteins of neutrophils (neutrophilic granule protein, NGP; neutrophil cytosol factor 2, NCF2; and neutrophil cytosol factor 4, NCF4) indicating a rapid migration of these cells into the wound area. This is supported by a higher abundance of integrin beta (ITGB2) and protein S100-A9 (S100-A9). Both proteins are involved in neutrophil migration towards wound sites Table 3. Identified proteins clustered to immune modulators. Identified proteins from proteomic analysis at 12 h after injury. Gene symbols and protein names correspond to Uniprot Knowledgebase. Quantitative values are given as Log10-transformed fold changes BD/STW. Significances are indicated as asterisks with * p-value < 0.05, ** p-value < 0.01 and *** p-value < 0.001. n = 4 for each defect scenario.  [70,71]. Neutrophils are known to be a key player in the early innate immune response [72]. Additionally, with the lymphocyte cytosolic protein 1 (LCP1) and coronin-1A (CORO1A) two proteins crucial for T cell activation and migration were found to be specific for the BD [73,74]. Moreover, annexin A1 (ANXA1) was upregulated in the BD group. This protein is involved in several pathways of the immune response but also has a crucial role for resolving the inflammatory response in later healing stages [19,75]. For the STW group only two proteins were found to be upregulated in the immune modulating protein cluster. Of those, the complement factor H-related protein (CFH) was shown to be a downregulator of the alternative complement activation [76].

Different Abundance of Proteases in Wound Fluids
One of the main functions of cells involved in the early immune response is the secretion of proteases either for defense, tissue remodeling or migration purposes. However, those proteases also lead to tissue damage, if the response is not resolved properly. Nine quantified proteins from the sampled wound fluids could be assigned to proteases (Table 4), including five of them being known to be secreted by neutrophils (cathepsin G, Cat-G; neutrophil elastase 2, EL; matrix metalloproteinase 8, MMP8; matrix metalloproteinase 9, MMP9; and proteinase 3, PR3) [77][78][79]. In inflammation the neutrophil proteases can activate or inactivate cytokines, activate specific receptors, and modify the activity of chemokines [80]. Five of nine proteins were unique or specific for the BD group, once more suggesting more intense immune response after a fracture as compared with a pure soft tissue wound. However, to the best of our knowledge, no influence on wound healing has yet been described for the BD-specific proteasome subunit beta type-10 (PSMB10) and napsin (NAPSA). Plasminogen (PLG) was the only protein identified in a higher abundance in the STW group. Besides its role in terminating blood coagulation it was shown to have a role in wound healing by remodeling of the ECM [81].

Different Abundance of Protease Inhibitors in Wound Fluids
A balanced activity of immune-related proteases is, among other mechanisms, achieved by the secretion of protease inhibitors. 20 proteins assigned to this cluster were quantified in the wound fluids, showing the importance of proteinase inhibitors in the wound area (Table 5). This includes nine proteins of the serine protease inhibitor protein superfamily (serpin) and four Inter-alpha-trypsin inhibitor heavy chains. Despite its structural similarities serpins inhibit a variety of serine proteases being involved in manifold biological processes, including tissue remodeling [19,82]. Inter-alpha-trypsin inhibitors are hyaluronic acid-binding proteins. By this binding they stabilize the ECM and prevent it from degradation [83]. Four wound scenario-specific proteins could be identified within this cluster. Serpin B1A -an inhibitor of the neutrophil-released Cat-G, EL and PR3 [84]-was the only protein with a higher abundance in the BD group. Alpha-1-inhibitor 3 (A1I3), its protein homologue murinoglobulin-1 (MUG1), and AMBP were specific for the STW group. A significant drop of both the mRNA expression in liver cells and the serum protein concentration of A1I3 could be shown within the first 24 h of the acute phase [85,86].

Different Abundance of Metabolites Elucidated by Metabolic Screening and Validated by Targeted Quantification
Metabolic screening and subsequent targeted validation revealed significantly different regulations of metabolite concentrations between BD and STW, as well as time-dependent differences. The largest differences between both wound fluids could be obtained by metabolic screening of the sampling interval 12-15 h (Fig 6AI-6AIV). This could also be verified by targeted approach indicated by highest amount of significantly regulated metabolites. Among other mechanisms, metabolic screening revealed significant time-dependent alterations of several amino acid pathways such as serine, tryptophan, histidine, arginine, and proline metabolism in BD and STW fluids. Targeted quantification of these potential key metabolites for wound fluid differentiation confirmed these findings with significant differences. In general most of the monitored amino acids in BD wound fluid showed a significant higher concentration at early time-point compared to STW wound fluid (Fig 6BI and 6BII). Since the investigated BD is the more severe type of inflammation this is in accordance to Yang et al. reporting elevated levels of e.g. serine in wound dialysates in early phase of inflammation as a function of immune response. After 24h elevated amino acid levels decreased to control level confirming our present data [87]. Another finding in metabolic screening was a variation of the carnitine shuttle. Quantification of 41 acylcarnitines in examined wound fluids revealed significant time-dependent differences between BD and STW samples (Fig 6BIII). Thereby it is noteworthy that concentrations of STW and BD wound fluids showed a similar appearance resulting in highest levels at 12-15 h followed by a significant decrease of carnitine concentrations and subsequent alignment to initial conditions. This course of acylcarnitine levels could be also explained by immune responses as also indicated by cytokine profiling (Fig 1) [88]. A differentiation of carnitine levels between STW and BD fluids was not obtained.

Conclusion and Outlook: Potentials of Combined Sampling of Wound Fluids
In this study we established a combined sampling by a minimally invasive technique with screening of two basic wound scenarios (bone defect and soft tissue wound) within the first 24 h after the injury. We were able to clearly differentiate between both wound scenarios and identify BD-specific features at all three levels of analysis, i.e. bone specific cytokine responses and distinct protein and metabolite profiles. Moreover, this differentiation is based on the capability to quantify cytokines, proteins and metabolites that can serve as a hallmark for a physiologic or dysregulated wound or fracture healing. The response of the acute phase mediator IL-6 was delayed in the samples from BD compared to STW. On the other hand, when using proteomics of the wound fluid, the negative acute phase markers AHSG and A1I3 were more abundant in STW. The neutrophil chemoattractants CXCL1, CXCL2 and CXCL3 were quantified with either a higher or earlier response in the BD dialysates. This finding was supported by wound fluid proteomics which revealed a higher abundance of neutrophil cytosolic proteins (NGP, NCF2 and NCF4) as well as by the presence of neutrophil migration factors (ITGB2 and S1009A). Moreover, neutrophil-released proteases (Cat-G, MMP8 and PR3) were enriched in the wound fluids of the BD group. With regard to an identification of a variety of protease inhibitors in both wound scenarios but with only one specific inhibitor for the BD group, a more intense innate immune response can be assumed after the femur fracture compared to the STW control group. This is supported by a higher abundance of fibrin and fibronectin constituting of a provisional ECM composed of fibrin and fibronectin, which were higher abundant in the STW group compared to the BD. Recently, non-healing wounds are mainly thought to be linked to an imbalanced and persistent immune response that overwhelms the tissue protective capabilities [19,72,77]. Hence, the monitoring of immune modulators, proteases and their corresponding inhibitors could serve as markers for the wound and fracture healing progress.
Future studies employing combined sampling should focus on the comparative monitoring of healing and non-healing bone defects under different conditions that could reveal the molecular background of the healing progress. Moreover, this technique could be used for assessing the host response towards different bone graft or bone substitute materials. However, this sampling approach is not limited to immune-related questions since proteins involved in several biological processes, like sugar metabolism and cytoskeleton organization, could be identified.
Supporting Information S1