Characterization of Regenerative Phenotype of Unrestricted Somatic Stem Cells (USSC) from Human Umbilical Cord Blood (hUCB) by Functional Secretome Analysis*

Stem cell transplantation is a promising therapeutic strategy to enhance axonal regeneration after spinal cord injury. Unrestricted somatic stem cells (USSC) isolated from human umbilical cord blood is an attractive stem cell population available at GMP grade without any ethical concerns. It has been shown that USSC transplantation into acute injured rat spinal cords leads to axonal regrowth and significant locomotor recovery, yet lacking cell replacement. Instead, USSC secrete trophic factors enhancing neurite growth of primary cortical neurons in vitro. Here, we applied a functional secretome approach characterizing proteins secreted by USSC for the first time and validated candidate neurite growth promoting factors using primary cortical neurons in vitro. By mass spectrometric analysis and exhaustive bioinformatic interrogation we identified 1156 proteins representing the secretome of USSC. Using Gene Ontology we revealed that USSC secretome contains proteins involved in a number of relevant biological processes of nerve regeneration such as cell adhesion, cell motion, blood vessel formation, cytoskeleton organization and extracellular matrix organization. We found for instance that 31 well-known neurite growth promoting factors like, e.g. neuronal growth regulator 1, NDNF, SPARC, and PEDF span the whole abundance range of USSC secretome. By the means of primary cortical neurons in vitro assays we verified SPARC and PEDF as significantly involved in USSC mediated neurite growth and therewith underline their role in improved locomotor recovery after transplantation. From our data we are convinced that USSC are a valuable tool in regenerative medicine as USSC's secretome contains a comprehensive network of trophic factors supporting nerve regeneration not only by a single process but also maintained its regenerative phenotype by a multitude of relevant biological processes.


Stem cell transplantation is a promising therapeutic strategy to enhance axonal regeneration after spinal cord injury. Unrestricted somatic stem cells (USSC) isolated from human umbilical cord blood is an attractive stem cell population available at GMP grade without any ethical concerns. It has been shown that USSC transplantation into acute injured rat spinal cords leads to axonal regrowth and significant locomotor recovery, yet lacking cell replacement. Instead, USSC secrete trophic factors enhancing neurite growth of primary cortical neurons in vitro.
Here, we applied a functional secretome approach characterizing proteins secreted by USSC for the first time and validated candidate neurite growth promoting factors using primary cortical neurons in vitro. By mass spectrometric analysis and exhaustive bioinformatic interrogation we identified 1156 proteins representing the secretome of USSC. Using Gene Ontology we revealed that USSC secretome contains proteins involved in a number of relevant biological processes of nerve regeneration such as cell adhesion, cell motion, blood vessel formation, cytoskeleton organization and extracellular matrix organization. We found for instance that 31 well-known neurite growth promoting factors like, e.g. neuronal growth regulator 1, NDNF, SPARC, and PEDF span the whole abundance range of USSC secretome. By the means of primary cortical neurons in vitro assays we verified SPARC and PEDF as significantly involved in USSC mediated neurite growth and therewith underline their role in improved locomotor recovery after transplantation. From our data we are convinced that USSC are a valuable tool in regenerative medicine as USSC's secretome contains a comprehensive network of trophic factors supporting nerve regeneration not only by a single process but also maintained its regenerative phenotype by a multitude of relevant biological processes. Injury to the spinal cord leads to a multiple damaging process including axonal contusion and transection with subsequent degeneration, massive apoptosis of oligodendrocytes and break-down of the blood-spinal cord barrier accompanied by invasion of immune cells resulting in sustained motoric and sensory impairments. Glial-fibrotic scarring and the lack of growth promoting factors impair axonal regrowth, which is currently the main target for therapeutic interventions to treat spinal cord injury. In addition, modulation of neuronal survival, remyelination of axons, and the immune reaction could promote functional regeneration (1)(2)(3). The inhibition of axonal regeneration might be overcome by exogenous application of growth factors or by transplantation of stem cells directly into the lesion site, which locally release trophic factors and thus support axonal regrowth. For clinical applications, stem cells should be ideally available on a clinical scale without ethical concerns or invasive interventions. Human umbilical cord blood (hUCB) 1 is an alternative stem cell source including cells similar to mesenchymal stem cells from bone marrow (BM-MSC) and can be easily expanded as adherent cells in vitro without any risk for the donor. Besides MSC, hUCB contains unrestricted somatic stem cells (USSC) (4), which can be clearly distinguished from BM-MSC as well as from hUCB derived MSC (CB-MSC) by their immunological behavior (5)(6), their transcriptome (7), the inability to differentiate into adipocytes (8), and by a specific Hoxgene expression pattern (9). Additionally, USSC exhibit a significant lower senescence rate and possess longer telomeres compared with CB-MSC and BM-MSC. As USSC can be easily isolated at GMP (good manufacturing practice) grade (10) and expanded on a clinical scale, USSC are a promising tool for transplantation studies. In a recent study, we have demonstrated that after transplantation into the acute injured rat spinal cord, USSC induce significant axon regrowth into the lesion side and effectively improve long-term functional locomotor recovery (11). Moreover, USSC transplantation promotes tissue sparing, which might contribute to the locomotor improvement. Although USSC were shown to differentiate into neuronal-like cells under in vitro conditions (4,12) replacement of endogenous cells in the injured spinal cord was not detected supporting the hypothesis that transplanted stem cells despite their lack of differentiation enhance regeneration by paracrine regulation or direct interactions with endogenous cells. In vitro studies confirmed the potential of USSC to promote neurite growth. Incubation of primary rat dorsal root ganglion neurons or cortical neurons with USSC conditioned medium (USSC-CM) significantly enhanced neurite growth, hence considering USSC-CM is an ideal tool to investigate USSC-derived neurite growth promoting factors. Kö gler et al. (13) provided first evidence using cytokine specific antibody arrays that USSC secrete axon growth promoting and neuroprotective factors in vitro such as stromal derived factor-1 (SDF-1) (14 -15), leukemia inhibitory factor (LIF) (16), and vascular endothelial growth factor (VEGF) (17)(18). In comparison, BM-MSC also secrete trophic factors, which enhance neurite growth in vitro, but the main trophic factors responsible for beneficial effects on neurite growth remain to be elucidated (19). Neurite growth promoting effects of BM-MSC seem to be at least partially mediated by brain derived neurotrophic factor (BDNF) and glial derived neu-rotrophic factor (GDNF) as demonstrated by neutralization of these neurotrophins using antibodies (20 -21). Classical neurotrophic factors, e.g. nerve growth factor (NGF), BDNF, and neurotrophin-3 (NT-3), have not been identified in the USSC secretome yet, and other factors released by USSC, which promote neurite outgrowth, have not been described in detail. Antibody array analysis of the secretome of other somatic stem cell populations of hUCB revealed that the cells release a large panel of cytokines and growth factors (22). Additionally, these cells express genes related to neurogenesis and blood vessel development as determined by gene expression profiling. For the unbiased identification of secreted proteins proteomic approaches including mass spectrometry is the method of choice. For example, CB-MSC secrete typical cartilage-related proteins during chondrogenic differentiation shown by nanoLC MALDI-TOF/TOF (23). Currently, the secretome of several stem cell types has been investigated by mass spectrometry (reviewed by (24 -25)). However, a secretome study of USSC derived from hUCB by LC-MS/MS focusing on nerve regeneration has not been performed, yet.
Here, we present the first study aiming at the characterization of the regeneration-supportive phenotype of USSC by detailed secretome analysis using protein mass spectrometry and exhaustive bioinformatic interrogation as well as subsequent functional validation of candidate neurite growth promoting proteins.
Protein Concentration-Different strategies have been used to concentrate proteins from USSC conditioned medium. Proteins have been (A) precipitated using a modified method published by Chevallet et al. (27), or (B) concentrated by lyophilization. Briefly, for protein precipitation (method A) 20 ml of conditioned medium was mixed with 10% sarkosyl NL (Sigma Aldrich). After adding trichloroacetic acid (Sigma Aldrich) to a final concentration of 7.5%, samples were incubated on ice for 2h with subsequent centrifugation at 10,000 ϫ g and 4°C for 10 min. The pellet was resuspended in 2 ml ice cooled tetrahydrofuran (THF, Fluka, Buchs, Switzerland). Afterward, samples were centrifuged again at 10,000 ϫ g and 4°C for 10 min. The pellet was washed carefully with 2 ml THF and dissolved in 100 l SDS sample buffer consisting of 600 mM DTT, 30% Glycerin, 12% SDS, 150 mM Tris/HCl, pH 7.0 (all Sigma Aldrich). After SDS-gel-electrophoresis (2 cm, 10 min) and silver staining according to Nesterenko et al. (28), the resulting lane was cut out in one slice. The gel slice was decolorized with a 1:1 mix of 30 mM sodium thiosulfate and 100 mM potassium hexacyanoferrate (III) (both Sigma Aldrich), washed, reduced (10 mM DTT) and alkylated (55 mM iodacetamide). Proteins were digested with 2 g trypsin (Serva, Heidelberg, Germany) overnight at 37°C. Resulting peptides were extracted with 50% acetonitrile (Biosolve, Valkenswaard, The Netherlands) and 0.05% TFA. For protein concentration by lyophilisation (method B) 20 ml of conditioned medium were lyophilized, reconstituted in 250 l SDS sample buffer and 10 l (2 g/l) was run into short SDS-gel, silver stained and subsequently digested and extracted as described above. Furthermore, a SDS-PAGE was performed and 12 lanes have been cut out for further analysis (method C). In supplemental Table S1 the methods are summarized.
LC-MS/MS-Peptides were separated using a UltiMate 3000 RSCLnano System (Dionex LC Packings (now Thermo Scientific)) with a C18-Pepman column (2 cm, 75 m I.D., 3 m particle size, 100 AA) as precolumn and a C18-Pepmap column (25 cm or 50 cm, 75 m I.D., 2 m particle size, 100 AA) as main column. A gradient of 0.1% FA (Fluka) to 0.1% FA/60% acetonitrile over 120 -150 min and a constant flow rate of 300 nl/min was used to elute peptides directly via electrospray (voltage 1.2-3.0 keV, capillary temperature 310°C) into the LTQ Orbitrap Velos or LTQ Orbitrap Elite mass spectrometer (both Thermo Fisher Scientific, Bremen, Germany). All MS spectra were recorded in positive ion mode with a mass range of 300 -2000 m/z and a resolution of 35.000 (Orbitrap Velos) or 350 -1700 m/z and a resolution of 60.000 (Orbitrap Elite). ϩ2, ϩ3 and higher charged (only Orbitrap Velos) monoisotopic precursors were isolated with a width of 2.0 Da and fragmented using a normalized collision energy of 35% for CID within the linear ion trap. A TOP20-data-dependent acquisition with polysiloxan as a lock mass was applied and dynamic exclusion activated (repeat count 1, duration 30 s on Orbitrap Velos and 45 s on Orbitrap Elite). Seven independent USSC secretomes were analyzed.
MS Data Analysis-Proteins were identified using Proteome Discoverer (Version 1.4, Thermo Fisher Scientific http://www.thermo scientific.com/en/product/proteome-discoverer-software.html) including Mascot search engine (Version 2.4.1, Matrix Science) (29). The UniProtKB/Swiss-Prot database (version from 2014/09, total entries: 546.439) was searched with a mass tolerance of 10 ppm (MSmode) and 0.4 Da (MS/MS mode), enzyme specificity was set to trypsin and two missed cleavage sites were considered during the search against human subdatabase. Mass range setting was 350 -5000 Da. Carbamidomethylation of cysteine was set as fixed modification. Oxidation of methionine was accepted as variable modification. Proteins were assembled by Proteome Discoverer with a Mascot threshold for protein identification set Ն 50.0 and one unique peptide.
For positive identification we considered a FDR of Ͻ 1% on peptide level (high peptide confidence, default p Ͻ 0.01). For FDR calculation we applied a decoy approach based on reversed protein sequences using Proteome Discoverer. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD002241 (30). Proteins were grouped by Gene ontology (GO) slim term cellular component and biological process.
Bioinformatic Analyses-The prediction of secreted proteins was performed with SecretomeP (31) as well as with SignalP (32). A NN-score Ͼ 0.6 (FDR Ͻ 0.05) was interpreted as a high probability of secretion via a non-classical pathway using Secretome P. We further applied functional annotation clustering to further characterize the USSC secretome using DAVID database (DAVID Bioinformatics Resources 6.7, (33), http://david.abcc.ncifcrf.gov/home.jsp) including Gene Ontology (GO) terms containing the sub-ontology biological process (defined by the Gene Ontology Consortium) with a false discoverer rate (FDR) Ͻ 0.05. Secretion of all UniProt listed proteins were predicted by SecretomeP and supplemented by proteins that are assigned to the UniProt term secreted or extracellular as well as classically secreted proteins with signal peptide predicted by SignalP. The 11.238 proteins served as background list for GO term annotation in DAVID database.
An absolute quantification of protein concentration was performed using Hi-N in Progenesis QI for proteomics 2.0 (Nonlinear Dynamics, Newcastle upon Tyne, see http://www.nonlinear.com/progenesis/ qi-for-proteomics/v2.0/faq/how-does-hi-n-work.aspx for further details) based on a publication by Silva et al. (34) using proteins concentrated by method A (TCA precipitation) and B (lyophilisation). Briefly, the Hi-N method considers all peptides for quantification. In the case if more than three peptides are available Hi-N includes the three most abundant unique (non-conflicting) peptides to calculate the absolute concentration. In contrast to top-3 method (34) the Hi-N algorithm applies the average and not the sum of the peptide intensities. Transferrin with a known concentration of 5 g/ml (64.8 pmol/ ml) was applied as single point calibrator and the protein signals were adjusted to transferrin for absolute quantification.
Preparation and Cultivation of Primary Rat Cortical Neurons-To test potential neurite growth promoting factors found in USSC secretome, primary cortical neurons were prepared from E15 Wistar rats as described before (11,35) and subsequently incubated with serumfree conditioned medium derived from USSC. N2 medium as well as conditioned medium derived from primary cortical astrocytes prepared from p0-p1 Wistar rats (11,36) served as negative and positive controls, respectively. Embryonic rat cortical neurons were seeded on 96 well plates (Costar, Corning, Wiesbaden, Germany) pre-coated with 1 mg/ml poly-D-lysine and 13 g/ml laminin (both Sigma Aldrich). A cell density of 40,000/cm 2 was proven to be optimal for automated neurite growth analyses. Cells were incubated at 37°C, 10% CO 2 and 98% humidity.
Neutralization and Stimulation Experiments-To influence activity of potential neurite growth promoting factors in USSC-CM, neutralizing antibodies have been incubated for 2 h at 37°C with USSC-CM prior to incubation of primary cortical neurons. Following antibodies have been used: anti-SPARC (secreted protein acidic and rich in cysteine, also known as Osteonectin, human, monoclonal produced in mouse, sc-33645, Santa Cruz Biotechnology, Heidelberg, Germany), 4 g/ml; anti-PEDF (pigment epithelium-derived factor, also known as serpinF1, human, polyclonal produced in rabbit, sc-25594, Santa Cruz Biotechnology), 4 g/ml; anti-CST3 (Cystatin C, human, polyclonal produced in goat, P01034, R&D System, Wiesbaden-Nordenstadt, Germany), 3 g/ml. Because antibodies have been solved in PBS, USSC-CM with appropriate amounts of PBS has been used as control.
Analyses of Neurite Growth-High content screening of cortical neurons was performed in vitro with a Thermo Scientific Cellomics® ArrayScan® VTI HCS Reader and the Neuronal Profiling v3.5 Bio-Application software (http://www.thermoscientific.com/en/products/ cellular-imaging-analysis.html). Plates were imaged with a 10ϫ objective (25 images per well) and simultaneously analyzed. For quantification, only Tub-positive structures with a DAPI-positive nucleus were included in cell counts. Following parameters were used for quantification: nucleus identification (iso data Ͼ Ϫ0.119; segment intensity Ͼ 80), nucleus validation (size 7-110), cell body identification (iso data Ͼ 0; minimum 1 cell body; seed segmentation; cell body segmentation ϭ 2), cell body validation (1 nucleus; cell body size 30 -1050; total cell intensity 20,450 -521,307), neurite identification (identification modifier ϭ Ϫ0.938; length 255; point resolution 2; gap tolerance ϭ 5), neurite validation (neurite length 10 -600; neurite intensity 0 -4095). Absolute numbers of neurons per well and total neurite length per neuron were used for analyses of different culture conditions. Significance of neurite growth was assessed by paired Student's t test comparing median neurite growth from six wells per condition and experiment with appropriate control condition. Neurite growth was considered to be significantly different at p Ͻ 0.05. All data are presented as a mean relative neurite growth (%) of at least three experiments Ϯ S.E.

Characterization of USSC Secretome by Mass
Spectrometry-The first evidence that the regenerative potential of USSC is likely to rely on paracrine mechanisms was obtained because we showed that USSC secrete factors that stimulate neurite growth of cultured dorsal root ganglion neurons and cortical neurons (11). As a first approach to test which molecular class mediates the trophic functions of USSC the supernatant was heated to 80°C for 5 min. The abolished neurite growth potential (data not shown) suggests that temperature-sensitive proteins are likely to enhance neurite growth. Therefore, we set-up a workflow analyzing serum-free supernatants of USSC using nano LC-ESI-MS/MS to identify proteins secreted by USSC. For protein preparation, we applied different methods to obtain a comprehensive protein catalogue of the USSC secretome (supplemental Table S1). Altogether, we identified 1520 proteins in USSC secretome (supplemental Table S2). Because of possible contaminations of conditioned medium by proteins originating from dead cells, not all identified proteins are likely to be secreted or extracellularly localized. Based on UniProt annotation, 993 proteins were classified as being potentially localized extracellularly and 118 proteins to be present at the cell surface (Fig. 1A, multiple protein entries possible). The USSC secretome further includes 884 membrane associated proteins, which could result from shedding. For further identification of secreted proteins, SignalP 4.0 algorithm was applied. We categorized 385 proteins to be classically secreted andusing SecretomeP 2.0 algorithm-276 proteins to be potentially secreted via non-classical pathways, remaining 487 proteins to be localized extracellular predicted by UniProt annotation (supplemental Table S2). Eight proteins, which have been found to be localized at the cell surface as identified by UniProt annotation, were not identified by SignalP and SecretomeP databases (Fig. 1B). Overall, 1156 proteins were determined to be candidate secretory proteins of USSC secretome, whereas 364 proteins were presumably contaminants from dead cells.
Clustering of USSC Secreted Proteins into Biological Processes-In order to gain an understanding of the complex composition of the USSC secretome, we assigned all proteins (1156 proteins) to biological processes by UniProt annotation resulting in biological processes associated with regeneration after spinal cord injury. USSC secreted proteins were as- ation (170 proteins), and cell growth (53 proteins) suggesting that USSC transplantation affects these biological processes relevant for nerve regeneration (Fig. 1C, multiple protein entries possible). Next, we performed enrichment analysis using DAVID bioinformatics database (DAVID Bioinformatics Resources 6.7, (33)) to reveal biological processes overrepresented in USSC secretome. Several biological processes were identified to be significantly enriched that are associated with neurite growth, neuronal differentiation and survival as well as nerve regeneration, e.g. cell adhesion (111 proteins), cell motion (58 proteins), cytoskeleton organization (42 proteins), and extracellular matrix organization (45 proteins) ( Fig. 2 and supplemental Table S3). Numerous proteins involved in cell adhesion expressed by USSC including cadherins, neuronal growth regulator 1, neuropilin 1, neuroplastin, roundabout axon guidance receptor and tenascin C are known to have a strong impact on nerve regeneration. Moreover, the USSC secretome included proteins associated with the func-tional category blood vessel development (44 proteins). The latter is known to be important for regeneration after neuronal trauma, containing angiopoietin 2, SDF-1, matrix metallopeptidases MMP-2, MMP-14, MMP-19, platelet-derived growth factor A (PDGFA), and TGF-␤2. In addition, the biological process collagen fibril organization (18 proteins) included proteins associated with extracellular matrix, e.g. lumican as well as aggrecan and TGF-␤2.
Quantification of USSC Secreted Proteins-Further we were interested to determine the concentration of identified proteins involved in relevant biological processes of neuronal regeneration. Therefore, we determined the protein concentrations based on a publication by Silva et al. (34) which enabled us to quantify 1020 proteins. For the USSC secretome we found that the protein concentrations varied in the range of 2-3 orders of magnitude with fibronectin as the highest abundant (88 pmol/ml) and quinone oxidoreductase (2 fmol/ml) as the lowest abundant protein detected by MS Transforming growth factor, beta-1 (TGF-b1) Growth factor signal peptide 45 P61812 Transforming growth factor, beta-2 (TGF-b2) Growth factor signal peptide 44 (supplemental Table S4). Quantified proteins were ranked into three classes based on their abundance. Enrichment analysis of the three abundance classes by DAVID bioinformatics database thereby using all secreted proteins as background revealed that the high abundant proteins (class I) are linked to extracellular matrix organization and ectoderm development (Fig. 3). Class II included proteins likewise associated with extracellular matrix organization but also with cytoskeleton organization and actin filament-based processes whereas class III proteins are linked to biological adhesion and cell adhesion. Analysis further revealed that 26 of the 31 identified neurite growth promoting factors were quantifiable and evenly distributed in class II (71-2457 fmol/ml) and class III (2-71 fmol/ml). Class II included growth factors and extracellular matrix (ECM) associated proteins whereas class III mainly included growth factors and small cytokine-like proteins. In class I, four neurite growth promoting proteins (2-88 pmol/ml) have been found that are all associated with ECM.

Validation of Potential Neurite Growth Promoting Factors in a Neurite Outgrowth Assay-
Neutralization of Candidate Neurite Growth Promoting Factors-With intend to validate neurite growth promoting function of potential candidates we investigated selected proteins assigned to first two abundance classes in a neurite outgrowth assay in vitro. Therefore, we inhibited functionality of candidate proteins by neutralizing antibodies applied to USSC-CM 2 h before application to primary rat cortical neurons. Cortical neurons were incubated for 48 h, fixed and stained with a neuron-specific antibody (␤-III-tubulin) to visualize cell bodies and neurites and analyzed by automated scanning with a Cellomics® ArrayScan® VTI HCS Reader. Astrocyte conditioned medium known to enhance neuronal survival and neurite growth in vitro (11,36,62) served as positive control for culture preparations in each experiment (data not shown). Our data indicate that neutralization of SPARC as well as PEDF (also known as serpinF1) in USSC-CM was effective with 4 g/ml neutralizing antibody each resulting in significantly decreased neurite growth com-pared with control cultures treated with USSC-CM (Fig. 4). Neurites were significantly shorter (67.6% Ϯ 2.3 as well as 74.8% Ϯ 4.1 of control) because of neutralization of SPARC and PEDF, respectively. Neutralizing antibodies against cystatin C (CST3) (Fig. 4) had no effect on neurite growth, which also indicates that application of immunoglobulin does not in general suppress neurite growth. Cell survival was unchanged upon neutralization of proteins (data not shown).

Stimulation of Neurite Growth by Application of Candidate Neurite Growth Promoting Factors-To further validate candidate proteins, we applied recombinant proteins in N2
(negative) control medium on primary cortical neurons and analyzed neurite growth after 48 h of incubation. Recombinant SPARC led to significantly enhanced neurite growth after application on cortical neurons up to 364.6% Ϯ 74.4 (Fig. 5). PEDF application alone had a small effect on neurite growth (139.7% Ϯ 18.6, p ϭ 0.07), which did not meet our significance criteria. Again, cell survival was not influenced by application of recombinant proteins (data not shown). In summary, neutralization of SPARC and PEDF in USSC-CM resulted in significantly decreased neurite growth. However, the decline in neurite growth did not reach the low basal levels of N2 control medium. We demonstrated further that application of recombinant SPARC to N2 medium enhanced neurite growth, but less strong than USSC conditioned medium. DISCUSSION USSC are a promising tool to support regeneration after spinal cord injury (SCI), because this stem cell type was previously shown to improve axonal regeneration and tissue sparing after transplantation into an acute rat SCI model (11). Beside axonal regeneration, support of cell survival and modulation of the ECM as well as angiogenesis are main targets for therapeutic interventions after SCI. As USSC secrete factors enhancing neurite growth and neuronal survival in vitro, we were interested to identify and characterize these factors in the USSC secretome to gain insight into potential paracrine mechanisms leading to en-

FIG. 2. Subset of biological processes associated with neurite growth, neuron differentiation, survival and regeneration enriched in comparison to all secreted proteins.
All proteins identified in the USSC secretome by LC-MS/MS excluding contaminants were included (multiple protein entries possible). GO terms were evaluated with DAVID database with a false discoverer rate (FDR) p Ͻ 0.05. hanced regeneration after USSC transplantation. In the present study we established a USSC secretome comprising 1156 proteins and revealed that USSC release at least 31 well-established neurite growth promoting factors. In addition, USSC release proteins that are related to several biological processes involved in nerve regeneration. We were able to functionally validate two of the candidate neurite growth promoting factors by neurite outgrowth assays in vitro validating our proteomic approach and confirming a paracrine mechanism of regeneration support for transplanted USSC in SCI.
Secretion of Proteins Involved in Neurite Growth or Axonal Regeneration-By means of secretome approach applying LC-MS/MS we determined 1156 proteins, whereof at least 31 proteins in the USSC secretome are already known to be associated with neurite growth or axonal regeneration. The group of 31 proteins include SPARC and thrombospondin-1 (TSP-1), which together with tenascin C modulate interactions with ECM molecules (63). We further identified netrin-4 known to attract or repel growing axons depending on neuron type (48). PDGF and TGF-␤1, both secreted by USSC, are known to interact with SPARC (64 -66), indicating a solid network between the secreted factors. Moreover, USSC secrete MIF, which directs the initial neurite outgrowth from the statoacoustic ganglion (SAG) to the developing inner ear (56), and NDNF, which could together with MANF, HDGF, and neudesin stimulate neurite growth as well as migration, cell growth, and survival (37)(38)(39)42). Recently, periostin, which is normally expressed in bone and muscles, has been shown to play an essential role in neurite growth and axonal regeneration after SCI to overcome the inhibitory effect of scar-associated molecules by signaling through focal adhesion kinase and Akt (67). Progranulin has similar characteristics like neurotrophins as it is cleaved into mature granulins by extracellular proteases and has been shown to stimulate neurite growth (68 -69). TGF-␤2, which is an immunomodulatory factor, normally facilitating ECM synthesis resulting in enhanced scar formation has also been shown to stimulate neurite growth in cultured dorsal root ganglion neurons (44). Using Hi-N method we showed that the identified proteins are distributed over a concentration range from fmol/ml -pmol/ml. As expected, growth factors were present at low concentration (fmol/ml range), whereas ECM associated proteins were highly abundant (pmol/ml range) in the USSC secretome. However, it has been shown that this kind of quantification method is associated with large quantification errors of small proteins (70), which likely depend on limited number of peptides as well as protein-to-protein variation. As our experiments were not designed for protein quantification in the first instance we took prior advantage of Hi-N method allowing quantification of hundreds of proteins in parallel to get a first impression about the distribution of proteins' concentration in USSC secretome. Although, this global approach gave acceptable results with mean standard deviation from 49 -115% concentrations of specific proteins need confirmation by complementary techniques.
Altogether, these different neurite growth promoting factors, which were expressed by USSC, either alone or in com-bination, are likely to participate in neurite growth stimulation of cultured primary cortical neurons as well as to support axonal regrowth observed after USSC transplantation into the injured spinal cord. The combined secretion of neurite growth promoting factors as well as inhibitory proteins led to en- hanced neurite growth as well as to axonal regeneration indicating an over-all supportive balance of USSC secreted proteins for regeneration associated processes.
Native USSC cultured in serum-containing medium were shown to release several cytokines and growth factors (13). In the present study, expression of CSF and SDF-1␣, which are known to be directly or indirectly involved in neurite growth or axonal regeneration (14 -15, 55), was confirmed under serumfree conditions. However, expression of the other proteins identified in the former work of Kö gler et al. (13) was not detected in our experiments, suggesting changes in the USSC expression profile under different culture conditions as well as the different detection modalities of the applied analytical platforms.
USSC secrete proteins involved in multiple regeneration associated processes-To establish the USSC secretome we performed rigorous bioinformatics data analysis. We considered accessible knowledge about proven extracellular localization and further we took advantage of the two bioinformatics tools SignalP and SecretomeP allowing prediction of classical and non-classical protein secretion, respectively. classical pathways. Three hundred and sixty-four proteins did not match our criteria and were removed from our candidate list. We are confident that our approach of combining experimental data with bioinformatics prediction tools is straightforward to establish USSC secretome. This holds true especially for classical protein secretion by signal peptide that is well studied and prediction algorithms like e.g. SignalP were successfully proven. In contrast, SecretomeP, which, based on simplified assumption that extracellular proteins share the same characteristics, have to be taken with caution and should only be applied in combination with experimental data.
Secretion of Proteins Involved in Extracellular Matrix Organization-Further data interrogation by annotation clustering revealed that USSC secrete proteins are involved in differentiation and developmental processes. For more detailed characterization of the USSC secretome, proteins were assigned GO terms by DAVID bioinformatics database revealing that USSC secrete proteins associated with biological processes involved in regeneration after SCI.
Several proteins of the USSC secretome were assigned to cell adhesion, extracellular structure organization, and extracellular matrix organization, e.g. the chondroitin sulfate proteoglycan (CSPG) aggrecan is known to inhibit axonal regrowth but also the growth promoting decorin, which antagonizes scar formation (46). USSC further secrete laminin subunits, which might promote neural differentiation and migration of neural precursor cells after CNS lesion and together with nidogen-1 could guide growth cone turning (71) as well as the heparan sulfate proteoglycan agrin, which is involved in synaptogenesis during regeneration (72). Thus, our results suggest that USSC strongly interact with the local environment after transplantation into the injured spinal cord via modifications of ECM structures as well as cell-cell-contact. Previous studies described that even in the presence of inhibitory molecules successful axon growth after injury were achieved when a prevalence of growth permissive signals is present (73)(74). Thus, USSC release ECM proteins that seem to shift the balance toward an axon growth permissive microenvironment thereby enhancing axonal regrowth in vivo.
Secretion of Proteins Involved in Angiogenesis-Beside axonal regrowth, regulation of cell adhesion and ECM modulation, angiogenesis is essential for proper regeneration as neurons have high metabolic requirements. Despite massive blood vessel sprouting, new blood vessels exhibit abnormal permeability after SCI suggesting limited functionality (75). USSC are likely to promote angiogenesis because these cells secrete several proteins assigned to blood vessel development or blood vessel morphogenesis, e.g. angiopoietin-2, angiopoietin-like 4, MMP-2, Ϫ14, and Ϫ19, PDGF, and TGF-␤, which together with guidance molecules such as netrin-4, semaphorin-7 and slit-3, could enhance blood vessel formation during regeneration.
USSC further secrete proteins grouped into GO terms cell motion and regulation of cell migration indicating that USSC might influence migration of endogenous neural precursor cells and endothelial as well as immune cells. Enhanced migration of precursor cells derived from the ependymal zone of the central canal could led to better recovery because these cells have been shown to differentiate into oligodendroglial cells (76). Because the SDF-1/CXCR4 axis is suggested to play a key role in migration and differentiation of adult endogenous stem cells and oligodendrocyte precursor cells after SCI (77)(78)(79), USSC expressed SDF-1 could influence precursor cell properties. In addition, secreted CSF-1, laminin subtypes, neuropilin-1, PDGF, migration and invasion enhancer 1 (MIEN-1) and TGF-␤ could regulate migration of precursor cells, astrocytes, fibroblasts, and immune cells, which contribute to proper regeneration.
Secretion of Proteins Involved in Cytoskeletal Organization-Most of the above mentioned proteins are known to be secreted via the classical pathway. In addition, the USSC secretome includes several proteins predicted to be either secreted by nonclassical pathways or annotated as extracellular by UniProt database. These proteins were found to be enriched in biological processes cytoskeleton organization, e.g. capping proteins, actin-related protein complexes, and gelsolin, which is known to play a role in neurite growth (80). Together with syntaxin, synaptotagmin, and kinesins, which were also found in the USSC secretome these proteins potentially result from exocytosis processes as these proteins are known to be involved in exocytosis. Furthermore, cytoskeleton proteins could originate from shedding because of their localization at the cell membrane, in particular ezrin and moesin both identified in the USSC secretome that link cell surface receptors and the actin cytoskeleton. It is unclear whether these proteins were secreted under certain conditions (25), but we cannot exclude that these proteins partially originate from dead cells. However, the biological processes related to the cytoskeleton included classically secreted proteins normally directly or indirectly regulating the dynamic remodeling of actin cytoskeleton (PDGF, SDF-1) indicating that USSC could also influence cytoskeleton dynamics of regenerating axons. By protein quantification we revealed that biological processes like, e.g. extracellular matrix organization and ectoderm development are enriched among highly abundant proteins, whereas proteins involved in cell adhesion are enriched within the group of low abundant proteins. We only can speculate if the abundance of a protein or a protein family is a measure for relevance of the associated biological process because both low abundant proteins like neurotrophins and high abundant ECM proteins are known to be critical for regeneration.
Functional Analysis of Candidate Proteins-SPARC is one of the most abundant proteins identified in the USSC secretome. It is a multifunctional glycoprotein that binds to collagens and vitronectins and modulates the activity of PDGF, FGF-2 and VEGF. It was further demonstrated that SPARC stimulates Schwann cells to facilitate axonal regeneration via laminin-1 and TGF-␤1 signaling (53). Schwann cells also express SPARC, which have been shown to promote both neurite growth and survival of purified retinal ganglion cells (81). SPARC is discussed to have synergistic effects on classical neurotrophins enhancing neurite outgrowth of retinal ganglion cells (82)(83). USSC-CM has been shown to stimulate neurite growth of primary cortical neurons (11). In the present study, we demonstrated that neurite growth is significantly decreased when a SPARC neutralizing antibody is applied to USSC-CM. On cortical neurons, recombinant SPARC directly enhances neurite growth. Therefore, we conclude that cortical neurons respond to SPARC signaling and that SPARC is one of the major neurite growth promoting factors secreted by USSC.
Recombinant PEDF (serpinF1) has been applied at low concentration representing the rather low amounts identified in USSC-CM, resulting only in a slightly increased neurite growth. On the other hand, neutralization resulted in a significant decrease in neurite growth, similar in range as SPARC antibody, indicating that PEDF may act in combination with other USSC secreted factors to stimulate cortical neurite growth. PEDF has been shown to enhance neuronal survival as well as neurite growth of retinoblastoma cells (84), cerebellar granule cells (85) and spinal motor neurons (86). Under our culture conditions, SPARC and PEDF did not affect neuronal survival of cortical neurons, indicating that other USSC secreted proteins promote neuron survival of cultured cortical neurons. In vivo, both proteins are likely to stimulate axonal regrowth directly but could also have indirect effects on regeneration via stimulation of glial cells or cell survival.

CONCLUSIONS
The lack of regeneration after SCI might be overcome by transplantation of stem cells directly into the lesion site, which locally release trophic factors and thus support axonal regrowth. The present study demonstrates that the regenerative phenotype of USSC secretome bases on a complex network of proteins associated with several biological processes that lead to enhanced neurite growth in vitro and therefore suggest a relevant role on axonal regeneration in vivo. The complex network of trophic factors secreted by USSC seems to act synergistically on several levels of nerve regeneration. With this observation it appears that transplantation of somatic stem cells instead or in combination with defined factors has great potential in regenerative medicine.