A new procedure for rapid, high yield purification of Type I collagen for tissue engineering

Process Biochemistry 44 (2009) 1200–1212

Contents lists available at ScienceDirect

Process Biochemistry journal homepage: www.elsevier.com/locate/procbio

A new procedure for rapid, high yield purification of Type I collagen for tissue engineering Xin Xiong b,1, Robin Ghosh c, Ekkehard Hiller a, Friedel Drepper d, Bettina Knapp d, Herwig Brunner b, Steffen Rupp a,* a

Fraunhofer Institute for Interfacial Engineering and Biotechnology, University Stuttgart, Nobelstrasse 12, 70569 Stuttgart, Germany Institute of Interfacial Engineering, University Stuttgart, Nobelstrasse 12, 70569 Stuttgart, Germany Department of Bioenergetics, Institute for Biology, University Stuttgart, Pfaffenwaldring 57, 70569 Stuttgart, Germany d Institute of Biology II, University Freiburg, Schaenzlestrasse 1,79104 Freiburg, Germany b c

A R T I C L E I N F O

A B S T R A C T

Article history: Received 27 April 2009 Received in revised form 4 June 2009 Accepted 19 June 2009

Collagen, which is one of the most abundant proteins in mammalian proteomes, is of significant medical and also biotechnological importance. In tissue engineering, collagen is used as natural matrix to facilitate growth of mammalian cells in 3D-structures. Here, large amounts of highly purified Type I collagen have been obtained from rat tail tendon by a two-step purification, suitable for industrial upscaling, involving extraction with 9 M urea followed by Superose 12 chromatography. Mass spectrometry identified only collagen Type I peptides indicating that the extracted collagen was homogeneous. Furthermore, a substantial amount of post-translational modifications were identified, including the presence of hydroxylated proline at the X-position in the G-X-Y repeats, which give new insight into collagen structure in vivo. The purified collagen was renatured quantitatively to form triplehelices by dialysis against water, as judged by UV-circular dichroism. This urea-extracted collagen (UC) was tested for its suitability as a matrix for growth of mammalian cells in tissue engineering. Cultures of a fibroblast cell line grown on urea-extracted collagen showed a higher motility and reduced stress levels than those grown on acetic acid-extracted collagen as determined by morphological studies and transcriptional analysis of selected marker genes using real time-PCR. These results indicate the suitability of the urea-extracted collagen obtained for biotechnological applications and tissue engineering. ß 2009 Elsevier Ltd. All rights reserved.

Keywords: Type I collagen Urea-extraction Mass spectrometry Self-assembly Cell culture Tissue engineering

1. Introduction Collagen is a major scaffold for biotechnological applications, e.g. in tissue engineering. This is one of the reasons why the molecular mechanism for the biosynthetic assembly of collagen is of high interest. Collagen is the major component in the extracellular matrix and Types I, II and III are the most abundant collagens which form fibrils responsible for tensile strength [1,2]. Due to its high accessibility and compatibility, Type I collagen is one of the most well-studied collagens and is already widely used as a bioscaffold in medicine and cell biology [3–6]. Type I collagen is trimeric [(a1)2a2] and exists as triple helix. There are several common sources for collagen: natural tissues such as skin and tendons [7,8], synthetic

* Corresponding author at: Fraunhofer-IGB, Nobelstr. 12, 70569 Stuttgart, Germany. Tel.: +49 711 970 4045; fax: +49 711 970 4200. E-mail addresses: [email protected] (X. Xiong), [email protected] (S. Rupp). 1 Present address: Institute of Experimental Internal Medicine, Medical Faculty, Otto von Guericke University, Leipziger Str. 44 39120 Magdeburg, Germany. 1359-5113/$ – see front matter ß 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.procbio.2009.06.010

collagen peptides [9,10] and cell cultures [11]. The spontaneous formation of triple helix from isolated a1 and a2 polypeptides in vitro has been studied extensively [12–15] and recent studies using synthetic model peptides [9,10] show the self-association or selfassembly of these peptides to a native-like triple-helix structure characteristically found in collagen fibrils. To isolate the collagen from cell culture or natural tissues the two most commonly used methods are neutral salt-extraction and low concentration acid-extraction using acetic acid and citric acid at a concentration of 0.05–0.5 M [16,17]. Reports that Type I collagen and procollagen have also been extracted using 8–10 M urea also exist [18,19]. However, this procedure has been described to yield only poor quality material, with significant degradation [18], making it non-attractive for further development. Collagen has been successfully applied for different purposes, e.g. as drug delivery system [20] and scaffold in tissue engineering [21,22]. In comparison to other biomaterials, naturally derived collagen shows excellent biocompatibility and safety due to its high abundance in all vertebrate animals and high biodegradability [23]. In addition, collagen also exhibits very low antigenicity [24].

X. Xiong et al. / Process Biochemistry 44 (2009) 1200–1212

Fibroblasts cultured using collagen matrix containing an ordered 3D-structure (3D cell derived matrix) take on a spindleshaped morphology, similar to in vivo fibroblast morphology. These cells show also increased proliferation, migration and adhesion compared to fibroblasts cultured on a collagen matrix without ordered structure (2D flattened cell-derived matrix) [25]. Binding of collagen to the integrins and other cellular receptors mediates cell adhesion and regulates the cell motility. Moreover, the interaction with proteoglycans and other components in extracellular matrix (ECM) can regulate the mechanical properties of different tissues [15,26]. Cell behavior can be affected by the changes of the ECM, in particular the fibrous collagen. It was shown that changes in the compression and alignments of the fibrils of collagen can lead to cellular signaling resulting in the formation of a regular geometric system pattern characteristic for the respective tissue [27]. Therefore, the matrix used in tissue engineering may be crucial for constructing a defined tissue- or organ-like structure. In this study we show that extraction of rat tail tendon with 9 M urea yields large quantities of homogeneous Type I collagen (a1, a2). The homogeneity of the preparation was confirmed by mass spectrometry (MS) and amino acid composition analysis. The morphology of fibroblasts grown on urea-extracted collagen (UC) and acetic acid/extracted collagen (AC) differed clearly. This was reflected in studies on a transcriptional level using qRT-PCR, whereby the regulation of Fak1, Rac1, MMP9 COL1A1 was determined. Our results indicate that purified urea-extracted collagen results in higher motility and reduced stress levels of fibroblasts than acetic acid-extracted collagen showing the suitability of urea-extracted collagen for biotechnological applications and tissue engineering. 2. Materials and methods 2.1. Protein isolation from rat tail tendons (RTTs) Rat tail tendon (RTT), teased out from 3-month-old rats was washed extensively in phosphate saline buffer (PBS) pH 7.4 at 4 8C. In order to get rid of traces of isocyanic acid in urea, the urea used was purified prior to extraction by an incubation with Biorad AG501-X8 resine (Nr.: 142424) for 30 min. The tendon was extracted with 9 M urea at 25 8C for 20 h, followed by a centrifugation at 4000  g for 30 min to sediment the insoluble components. The urea-soluble proteins were purified by gel-filtration using a Superose 12 column (GE-Healthcare, diameter 6 cm and height 60 cm) at a flow rate of 25 ml/h. The column was equilibrated with running buffer, containing 8 M urea, 50 mM Tris–HCl pH 7.5, and 100 mM NaCl. The change of urea concentration from 9 M to 8 M is necessary to prevent the precipitation of urea during the gel-filtration. Collected fractions were then dialyzed repeatedly against a 100-fold volume of distilled water at 4 8C for two days and then lyophilized. For further analysis the urea-soluble proteins were fractionated by using a SMART-system (Amersham Biosciences) through a smaller Superose 12 column (GE-Healthcare, diameter 3.2 mm and height 300 mm). The proteins were detected by monitoring the UV absorbance at 213 nm, 256 nm and 280 nm. For acetic acid extraction, the PBS washed rat tail tendon was incubated in 0.5 M acetic acid and stirred at 4 8C for 7 days, followed by a centrifugation at 4000  g for 30 min to sediment the insoluble components. The clear supernatant was then stored at 4 8C [17]. 2.2. Amino acid composition analysis The UC and AC samples were hydrolyzed with 6 M HCl in the absence of oxygen at 110 8C for 24 h. The amino acid composition of UC was determined on an ICS 3000 (Dionex) instrument using an Amino PAC-PA-10 column (with pre-column) according the manufacturer’s instruction. The composition of AC was determined using a HPLC-system (Agilent Technologies) with pre-column derivatization according manufacturer’s instruction. The quantitation of amino acids was performed using a standard mixture (Catalogue Nr.: 5061-3330 and 5062-2478 Agilent Technologies. The concentration of single amino acids was 1 nmol/ml in 0.1 M HCl). 2.3. Cyanogen bromide cleavage Lyophilized urea-extracted collagen (UC) from gel-filtration was dissolved in formic acid containing 30% ethanol (v/v) at a collagen concentration of 3 mM as determined by the DA220 (see 2.8). For the cleavage of methionine, cyanogen

1201

bromide (CNBr) 75-fold in excess of the His amount which can be estimated from the DA220 (unpublished data) was added to the solution in a Reacti-Vial (Pierce) and the reaction was performed for 6 h in the dark at room temperature under stirring. The reaction mix was centrifuged at 16,000  g for 10 min and the supernatant was transferred to an Eppendorf tube for drying in a Speed-Vac (Univapo-100H, Uniequip GmbH, Germany) for 2 h. 2.4. Carboxypeptidase digestion Dialyzed UC and AC samples were digested with sequencing grade yeast carboxypeptidase Y (Roche Applied Bioscience, Germany, Catalogue Nr. 1111914001, sequencing grade, analyzed to be free of other proteolytic activity. Purity was proved by amino acid composition analysis and SDS-PAGE.) for 1 min to 20 h with enzyme/collagen (E/C) ratios varied from 1/200 to 1/25 (w/w). The E/C ratio was also increased to 1/50 and 1/25 for 20 h. The digestions were carried out at 37 8C and stopped by the addition of PMSF. The digested samples were then reduced and analyzed by 6% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). 2.5. Electrophoretic analysis Routinely, SDS-PAGE was performed according to Laemmli [28]. For SDS-PAGE analysis, lyophilized fractions were dissolved in sample buffer, containing 1% SDS, 5% b-mercaptoethanol, 25 mM Tris–HCl pH 7.5, 1% glycerol and 0.05% bromophenol blue, and then boiled at 100 8C for 5 min prior to gel loading. The gels were usually stained with Coomassie Brilliant Blue R250 using a standard protocol. 2.6. Mass determination To determine the mass of the collagen protein, a thin-layer preparation with a sinapic acid matrix was used in MALDI-TOF MS. 1 ml sinapic acid (20 mg/ml in acetone) was spotted on a stainless steel MALDI target (Bruker Daltonik GmbH, Bremen, Germany) and dried. 1 ml aqueous protein solution (2.5 mg/ml) was deposited on the matrix layer and also dried. Afterwards, the sample was covered with 1 ml sinapic acid solution (20 mg/ml) dissolved in 0.1% aqueous trifluoroacetic acid containing 40% acetonitrile and incubated to dryness. The average mass of the proteins was analyzed by MALDI-TOF MS using an Ultraflex II TOF/TOF 200 (Bruker). The mass spectrometer was set to scan over a mass range of 10–140 kDa. For primary analysis, scan was extended over a range of 10–320 kDa. The mass spectra were acquired in linear mode and processed using FlexControl and FlexAnalysis Software (Bruker). b-galactosidase monomer (116,483 Da; Calbiochem) and glycogen phosphorylase (97,289 Da; Sigma–Aldrich) were used for external calibration. 2.7. Tandem mass spectrometric (MS/MS) analysis About 0.5 mg/ml of protein was dissolved in a solution containing 8 M urea and 50 mM NH4HCO3, then reduced and alkylated using standard protocols. The sample was diluted and tryptic digestion was carried out at 37 8C over night. After digestion the solution was adjusted to a pH of 2–3 with concentrated formic acid. Peptide mixtures were separated for nano-LC-ESI-MS/MS using a FAMOS autosampler (Dionex), an Ultimate inert HPLC (Dionex) and an Agilent HPLC 1100 pump connected to the nano-ESI-Source of a Finnigan LTQ-FT (Thermo Electron Corporation) for online mass detection. Peptides were first collected on a trap column (dimension 0.1 mm  15 mm, self-made using Zorbax Eclipse XDB-C18, 5 mm, Agilent Technology) for desalting and concentrating followed by separation on an analytical column which was self-made with loader kits SP035 from Proxeon Biosystems using Hydrosphere C18, 3 mm (YMC) and fused silica emitters (0.075 mm  105 mm, 6 mm, Proxeon Biosystems). Peptides were eluted during a 60 min gradient using 97% water, 3% acetonitrile and 0.1% formic acid as solvent A and 80% acetonitrile, 20% water and 0.1% formic acid as solvent B at a flow rate of 0.15 mL/ min. Mass spectrometric detection consisted of FT full scans at a resolution of 25,000 followed by data-dependent FT SIM scans at a resolution of 50,000 and low resolution IT MS/MS scans using a dynamic exclusion of parent ion masses for 60 s. The MS and MS/MS spectra were submitted to an in-house installation of the open mass spectrometry search algorithm [29] (OMSSA, version 2.1, NCBI) and searched against NCBI non-redundant (nr) database (release February 8, 2007) and its subsets. Search results were filtered and sorted using in-house written software (F. Drepper, unpublished data). Peptide hits were considered significant if the precursor and product ion masses matched within 3 ppm and 0.5 rel. mass units, respectively, and if the E-value was below 0.01. 2.8. Spectroscopic analysis UV-spectra were obtained using a JASCO V-560 spectrometer using a scanning speed of 200 nm/min and slit width of 2 nm. Generally, 2 mm path-length quartz cuvettes were used for the measurements. UV–CD spectra were acquired with a JASCO J-715 CD spectrometer (scanning speed of 50 nm/min, 2 mm cuvette, slit width 2 nm, averaging over 9 scans). Spectra were obtained in the range of 320 nm to 190 nm. For the spectroscopic measurements, all buffers were degassed under vacuum prior to use. All spectra were taken at 25 8C.

X. Xiong et al. / Process Biochemistry 44 (2009) 1200–1212

1202 2.9. Reverse-phase-HPLC (RP-HPLC)

3. Results

For separation and analysis of the CNBr-cleaved peptides, RP-HPLC was performed with an SMART system, using a C18-column 30 nm column. Peptides were applied to the column using 0.1% trifluoroacetic acid [30] (v/v) in H2O and then eluted with a gradient containing 7–40% acetonitrile/0.1% TFA (v/v). The flow rate was 0.1 ml/min. Peptides were monitored with three different wavelengths (215 nm, 256 nm and 280 nm). Prior to loading to the column, the dried CNBrcleaved peptides were dissolved in aqueous solution containing 0.1% (v/v) TFA and 5% acetonitrile. 2.10. Gaussian analysis of UV–vis spectra Gaussian deconvolution [31] of UV–vis spectra was performed using an in-house program using the following algorithm: (1) the baseline absorbance due to turbidity at the red end (300 nm) of the spectrum was subtracted from the complete wavelength region; (2) a wavelength-dependent correction (K/l4, where K is an empirical constant) was then applied. The value of the K-constant was adjusted to reduce the blue end (260 nm) of the spectrum to those of reference solution spectra of tyrosine (Tyr) and tryptophan (Trp); (3) Gaussian curves were then fitted to the baseline-corrected spectrum until the sum of residuals (experimental-minuscalculated) was less than about 1%. 2.11. Cell culture and coating of cover-slips using collagen The NIH 3T3 fibroblasts were grown in Dulbecco’s modified Eagle medium (DMEM) (GIBCO) in a tissue culture flask (Greiner) at 37 8C, 5% CO2, until reaching confluence. This DMEM contains 10% fetal calf serum (FCS) (Clonetics), 2 mM Lglutamine, 0.1% gentamycin, 0.45% glucose and 0.1% penicillin–streptomycin. The same amount of UC and AC were mixed with DMEM (1 mg/ml) and rapidly filled into 24 well-plate with inserts for expression analysis and into 6-well plate with cover-slips for phenotypical studies by SEM. The collagen solutions were placed under a clean bench for 2 h to gel the collagen. Then, 104 cells in 100 ml DMEM were dropped onto the collagen surface in the inserts. For the collagen-coated cover-slips 5000 cells were seeded. The medium was changed every two days. 2.12. Scanning electron microscopy (SEM) NIH 3T3 fibroblasts were cultivated routinely as described [32]. For morphology studies, 5000 cells were seeded to the collagen-coated glass slides and cultivated for one week. Medium was removed; samples were washed twice in PBS buffer and fixed for 30 min at 4 8C with 2.5% glutaraldehyde in PBS buffer, pH 7.4). Following fixation, samples were treated for 30 min with a 2% osmium tetroxide solution in PBS buffer. Samples were then dehydrated with graded ethanol (from 50% to 100%) and eventually were sputtered according to standard protocols. The SEM-imaging was performed by using a field emission electron microscope (FEEM) (Model 1530 VP, Leo Electron Microscopy Ltd., US). 2.13. RNA-preparation and qRT-PCR NIH 3T3 fibrobalsts were grown on different surfaces including glass, UC and AC for 14 days with a medium change every 2 days. The cultures were then frozen in liquid nitrogen for generating cell pearls. Cell pearls were then disrupted mechanically by grinding in a pre-cooled Retsch-mill (Retsch, Haan, Germany) and RNA was further purified using the RNeasy Mini Kit (Qiagen, Hilden, Germany) following instructions of the manufacturer. For the qRT-PCR analysis, genes involved in regulation of cell motion, matrix remodeling and mechanical stress response were tested. Amounts of cDNA of these selected genes of interest in NIH 3T3 fibroblast grown on UC were compared to those on AC. Glyceraldehyde-3-phosphate dehydrogenase (G3P) was chosen as reference gene with the assumption that the transcript amount of this gene in any culture under any condition is approximately equal. The transcript amounts from genes of interest were compared to these in reference culture on glass and normalized by comparison of the transcript amounts of the reference gene in the samples. The real-time PCR was performed using a LightCycler 480 (Roche, Germany) with LightCycler 480 master kit following manufacture’s instruction. The gene specific primers were designed according to suitable UPL-probes (Roche, Germany) and purchased from TIB MOLBIO, Germany. The whole primer list and UPL-probes were given as follows: Gene

UPL Nr.

Forward primer

Reverse primer

G3P Actin b Rac1 Rha Fak1 COL1A1 TGF-b1 MMP9

29 64 77 92 45 15 72 19

50 -gagccaaacgggtcatca-30 50 -ctaaggccaaccgtgaaaag-30 50 -agatgcaggccatcaagtgt-30 50 -gaatgacgagcacacgagac-30 50 -gctacaatgagggtgtcaagc-30 50 -atgttcagctttgtggacctc-30 50 -tggagcaacatgtggaactc-30 50 -acgacatagacggcatcca-30

50 -catatttctcgtggttcacacc-30 50 -accagaggcatacagggaca-30 50 -gagcaggcaggttttaccaa-30 50 -tcctgtttgccatatctctgc-30 50 -ggtcaaggttggcagtgg-30 50 -gcagctgacttcagggatgt-30 50 -gtcagcagccggttacca-30 50 -gctgtggttcagttgtggtg-30

3.1. Collagen extraction and SDS-PAGE analysis Collagen samples obtained from rat tail tendon of 3-month-old rats after extraction with 9 M urea following our procedure yielded a clear solution. The yield of the urea-extraction without purification via gel-filtration was calculated from three isolations using different amounts of rat tail tendons. To determine the yield, the tendons were first washed in PBS and water then lyophilized. The dry weight of the lyophilized rat tail tendon was determined and after 20 h incubation in urea (see Section 2) the obtained solution was centrifuged and the supernatant was removed and dialyzed against water. Determination of the dry weight of every fraction from the extraction process yielded up to 95% of ureaextracted collagen (UC) (Table 1). The lyophilized tendons could be easily dissolved in 9 M urea and showed a clear solution after the centrifugation. Interestingly, collagen extracted using acetic acid (according to [8,17]) precipitates in 9 M urea and cannot be dissolved for further purification, indicating changes in structure of AC. SDS-PAGE analysis of the UC, performed under reducing conditions showed a complicated protein pattern, with many bands above the expected molecular weights of collagen a1 and a2 of about 130 kDa (a1) and 140 kDa (a2), respectively (Fig. 1B, lane 1) without any detectable degradation products. Most likely, due to the structure of the collagen or due to post-translational modification, the relative migration in polyacrylamide gels results in an about 20% increase of the observed mass in comparison to that of the predicted mass [33]. No Coomassie-stained band below 100 kDa was observed. The extracted collagen was then purified further for more detailed characterization using gel-filtration according to the procedure described in detail in methods. Purification of the urea-extracted samples by Superose 12 gelfiltration in 8 M urea yielded an unexpected profile (Fig. 1A) showing three overlapping peaks at an elution volume where high molecular weight species were expected. SDS-PAGE of the higher molecular weight fractions showed that the sub-peak I (Fig. 1A) corresponded predominantly to aggregates of an apparent high molecular weight >300 kDa but also contained lower molecular weight aggregates found predominantly in sub-peak II (Fig. 1B). However, sub-peak II also contained species with molecular weights corresponding to a1 and a2 which dominated the SDSPAGE profile found in sub-peak III. We note that the aggregates observed by SDS-PAGE cannot be accounted for by disulfide bridge formation as samples were routinely reduced with 5% b-ME prior to loading. Also, SDS-PAGE profiles obtained with thiolglycolic acid in the running buffer were identical to those shown in Fig. 1 (data Table 1 Yield of the urea extraction. Isolation

RTT dry weight (g)

PBS soluble dry weight (g)

Insoluble dry weight (g)

Yield (%)

1 2 3

0.1955 0.2644 0.3202

0.0048 0.0037 0.0085

0.0047 0.0146 0.0167

95.1 93.1 92.1

The PBS-soluble fraction is the fraction that is removed after the washing of rat tail tendons with PBS. The insoluble fraction is the remaining precipitate after urea-extraction. The yield is equal to the (dry weight of rat tail tendon) minus (dry weight of insoluble fraction in urea and removed fraction by PBS-washing) than divided by (dry weight of rat tail tendon). An average collagen extraction yield was performed in triplicate using the calculation: Y UE ¼ ððW RTT  W PBS  W I Þ=W RTT Þ  100%. YUE is yield of urea-extraction. WRTT is dry weight of rat tail tendon. WPBS is the dry weight of components remaining in PBS by the wash steps. WI is the dry weight of urea-insoluble fractions.

X. Xiong et al. / Process Biochemistry 44 (2009) 1200–1212

1203

Fig. 1. Superose 12 chromatography and SDS-PAGE analysis of urea-extracted collagen. (A) Chromatogram of the gel-filtration profile following the initial urea extraction (first run). The black bar indicates the fractions which were pooled prior to re-chromatography using the same column (second run shown in (C and D)). (B) SDS-PAGE profile of the fractions shown in (A). Lanes 2–7 corresponds to sub-peak I, lanes 8–12 corresponds to sub-Peak II and lanes 13–16 corresponds to sub-peak III. (C) Re-chromatography of fractions 8–12 taken the first run and the corresponding to SDS-PAGE profile (D). (D) SDS-PAGE profile of the fractions shown in (C). Lanes 1–5 corresponds to sub-peak I, lanes 6–11 corresponds to sub-peak II and lanes 12–16 corresponds to sub-peak III. The arrows shown in panels A and C indicated the sub-peaks (I, II and III) mentioned in the main text. 5% SDS-PAGE was used.

not shown). To our surprise, re-chromatography of each of the subpeaks on the same column under the same conditions yielded an almost identical elution and SDS-PAGE profile. We observed, however, a bias to high or low molecular weight species in dependence of the protein-concentration. Re-chromatography of the pooled fractions from sub-peak II (predominantly species with molecular weight of about 300 kDa) consequently yielded a larger fraction of monomeric a1 and a2 than the SDS-PAGE profile of the original pool suggested (Fig. 1C and D). A third chromatographic step of any sub-peak also yielded the characteristic threecomponent profile (data not shown). The distribution of the apparent molecular weight of the purified collagen is biased to decreasing molecular weight with each subsequent gel-filtration run. This is the opposite to what would be expected for aggregation of polypeptides due to cross-linking or non-specific hydrophobic interactions. Thus, the chromatographic behavior of the ureaextracted samples seemed to indicate a reversible association of collagen in 8 M urea. Moreover, dialysis of any of the Superose subpeaks I, II and III against H2O yielded a sol–gel preparation [34,35], completely lacking any precipitates. Cross-linking of a1 and a2 chains could also be excluded due to the detection of carbamylated Lys (about 25% of the total) in the collagen chains, which may have taken place during the urea-extraction and were detected using ESI-MS/MS. All Lys residues present in the SwissProt-sequences were detected in our MS/MS analysis indicating that the majority of the Lys residues are most probably not cross-linked. The total recovery of the urea-procedure after the gel-filtration was also determined as we pooled the major fractions from 95 collected fractions (35 fractions, 10 ml/fraction) and lyophilized the fractions after extensive dialysis against water. The recovery of purified collagen was then calculated to be about 62%-65% (g recovered collagen/g RTT) of the starting material. 3.2. Mass determination and tandem MS/MS The total mass of the isolated collagen species from all three sub-peaks of the gel-filtration were determined by MALDI-TOF MS using b-galactosidase (monomer 116.483 kDa, SwissProt Nr.:

P00722) and glycogen phosphorylase B (97.289 kDa sp: P00489) as standards. The mass spectrum of these fractions is shown in the Fig. 2A. Interestingly, all fractions have approximately the same molecular weight of about 96 kDa within the accuracy-range of 20 ppm, strongly arguing for only one protein species present in the sample. The peaks for two- and three-fold charged collagen peptides were also observed in this spectrum (Fig. 2A). The mass of acetic acid-extracted collagen (AC) was also measured (Fig. 2B.) Interestingly, we detected a mass difference of up to 3 kDa. In particular, urea-extracted collagen showed 2 peaks at 95.942 kDa and 94.556 kDa while the acetic acid-extracted collagen has only one signal at 93.559 kDa. The two-fold charged peaks showed the same patterns (Fig. 2B). Scans in a high molecular weight range up to 300 kDa showed no detectable peak. We also incubated the native rat tail tendons in sinapic acid within a time range from 10 min to 60 h and analyzed the preparations using MALDI-TOF as described for mass determination (see Material and methods 4.5). We did not observe any increase of the signal intensity in mass spectra indicating that, sinapic acid has no effect on the integrity of collagen (Supplementary Fig. S1). Even after 60 h the RTT were not significantly solubilized in sinapic acid. Cross-linked species of collagen could not be detected. For protein identification, tandem MS analysis was performed as described in experimental procedures. The isolated and purified collagen was digested using either trypsin or CNBr as described in experimental procedures. Nano-LC-ESI MS/MS was first applied to analyze the tryptic-digested collagen samples. Fractions representing the three sub-peaks detected after chromatography (shown in Fig. 1) were analyzed separately and mass lists were then submitted to database search. Again, the protein content of all three fractions was identical. Only collagen a1 and a2 chains were detected which contained only peptides corresponding to rat Type I collagen with a coverage for collagen a1 of 95.8%, (shown in Fig. 3A) and for a2 sequences of 88.1% (Supplementary Fig. S2) (sp: P02454, P02466). In parallel, AC was also analyzed by ESI-MS/MS. However, using AC, only 60% sequence-coverage (66.5% for a1 and 60.8% for a2) was obtained (Supplementary Fig. S3A and B). This is presumably due to the lowered integrity of AC after the acetic acid-

1204

X. Xiong et al. / Process Biochemistry 44 (2009) 1200–1212

Fig. 2. Mass determination of UC and AC. (A) Mass determination and comparison of fractions from sub-peaks as shown in Fig 1A. (B) Mass comparison of urea-extracted collagen and the acetic acid-extracted collagen. Inserts are enlarged spectra of one-fold charged and two-fold charged collagen. The black highlighted spectrum is from UC and grey spectrum is from AC.

extraction which may also explain the mass difference of 3 kDa to UC (Fig. 2B). These results were confirmed by MALDI-TOF/TOF using MASCOT for protein database search. Also automatic de novo based MS/MS protein identification using PEAKS showed the same identification- and sequence-results [36]. Furthermore, amino acid composition analysis showed that AC had a significantly lowered Gly, Pro as well as Hyp contents as compared to UC and the theoretical values (Fig. 3B and Supplementary Table S1). This could be due to the impurities remaining after the standard acetic acid extraction procedure. In contrast to AC samples, in UC we identified 42.6% of the total amino acids as Gly, 10.5% as Hyp and 13.2% as Pro, i.e. 44% of the Pro are hydroxylated. The amino acid composition obtained from UC matched the theoretical values of SwissProt-sequences (Gly 33%, Hyp + Pro 21.47%) with only 6.7% error. In particular, the Pro and Hyp determinations are in good agreement with the data from ESI-MS/MS results (Supplementary

data). 92–120 proline residues from 236 Pro in a1 chain and 70–90 Pro from 201 Pro in a2 chain were detected as Hyp in our UC samples (Hyp content varies between 37% and 48% as determined by MS/MS). In contrast to UC, in AC samples, rat albumin, propeptides of Type I procollagen and other proteins could also be detected with high significance, indicating that gel-filtration has removed these impurities. Post-translational modifications (PTM) are thought to be crucial for formation of the tertiary and quaternary structure of collagen. In particular, the hydroxylation of Pro and Lys residues are very important for the folding and stability of folded triple helices. Therefore, we determined hydroxylation of Pro and Lys present in purified urea-extracted collagen (UC) using tandem MS analysis. Surprisingly, some peptides with the same primary sequence were found to be differentially modified with regard to Pro- and Lys-hydroxylation (Tables 2 and 3). Moreover, some Hyp

X. Xiong et al. / Process Biochemistry 44 (2009) 1200–1212

1205

Fig. 3. ESI-MS/MS sequence coverage and amino acid composition analysis of UC. (A) Tandem MS results. The bold letters show the identified sequences. (red: detected in CNBr-cleaved fractions only). Residues differing from SwissProt entry P02454 are underlined in green. The sequences in italics prior and behind the core region are the propeptides sequences, which should be not presented in the mature collagen. The maximal amount of detected Hyp is indicated in the first column on the left site of the figure and the minimal amount of Hyp is denoted in brackets. The numbers of hydroxylysines are indicated in the second column as shown for Hyp. (B) Comparison of amino acid composition between theoretical values (SwissProt P02454 and P02466) UC as well as AC sample. Theoretical values are in blue, values of UC are in purple and AC in yellow. The content of Hyp is shown on the proline column as indicated by black bar.

were identified at unusual X-position but not at the generally accepted Y-position in the G-X-Y sequence (Tables 2 and 3). In parallel, the AC was also analyzed with regard to the differential hydroxylation of Pro (Supplementary Table S2 and S3). Due to the low coverage of sequence, an absolute comparison between UC and AC was not possible; however, the differential hydroxylation patterns in the peptides identified are the same in both UC and AC samples. 3.3. UV–CD- and UV–vis-spectroscopy UV–CD and UV–vis spectroscopy were employed to characterize the physical state as well as possible conformation changes of the purified urea-extracted collagen under a variety of conditions. UV–CD spectra (Fig. 4) of the UC sol–gel showed a characteristic peak at 225 nm, attributable to the presence of triple helices [9].

We could demonstrate a perfect correspondence between the UV– CD data, obtained from synthetic polypeptides (SC) which form triple helices exclusively [9], and our data obtained from the sol– gel preparation of urea-purified collagen (Fig. 4A, red line). The fit with the published data is shown in Fig. 4B. The match of both spectra is about 98%. The UV–CD spectra from UC were compared to those literature data obtained from calf skin AC, which were considered to be ‘‘native’’ collagen spectra. However, this calf skin AC spectrum shows a slight blue shift if compared to both UC and SC spectra, indicating that this material may also contain random coils (Supplementary Fig. S4). In order to examine the mechanism of collagen assembly from the urea-purified polypeptides in more detail, we measured the UV–CD spectra for collagen polypeptides dissolved in various concentrations of urea (Fig. 4A). In 8 M urea, the UV–CD spectra are characteristic of random coil as expected. Essentially, no change in

X. Xiong et al. / Process Biochemistry 44 (2009) 1200–1212

1206

Table 2 Differential hydroxylation of proline residues in Col 1a1 chain. Sequencea e

Position

Mrb

Ionsc

E-valued

1a

DGLNGLPGPIGPPGPR DGLNGLPGPIGPPGPR DGLNGLPGPIGPPGPR

1142–1157

1560.7895

14 13 13

3.1E04 8.3E04 8.3E04

1b

DGLNGLPGPIGPPGPRe,f DGLNGLPGPIGPPGPRe,f DGLNGLPGPIGPPGPRe,f

1142–1157

1544.7949

13 14 14

1.9E04 3.2E04 3.2E04

2a

TGPPGPAGQDGRPGPAGPPGAR TGPPGPAGQDGRPGPAGPPGAR TGPPGPAGQDGRPGPAGPPGARe TGPPGPAGQDGRPGPAGPPGAR TGPPGPAGQDGRPGPAGPPGAR

542–563

2029.9571

24 24 25 25 25

2.6E08 3.8E08 1.9E07 2.7E07 5.2E07

2b

TGPPGPAGQDGRPGPAGPPGARe,f TGPPGPAGQDGRPGPAGPPGARe,f TGPPGPAGQDGRPGPAGPPGAR TGPPGPAGQDGRPGPAGPPGAR TGPPGPAGQDGRPGPAGPPGAR TGPPGPAGQDGRPGPAGPPGAR

542–563

2013.9623

33 33 31 31 31 31

1.3E15 6.4E15 6.0E14 3.0E13 3.7E13 1.5E12

3a

GETGPAGRPGEVGPPGPPGPAGEKe,f GETGPAGRPGEVGPPGPPGPAGEKf GETGPAGRPGEVGPPGPPGPAGEK GETGPAGRPGEVGPPGPPGPAGEK GETGPAGRPGEVGPPGPPGPAGEK GETGPAGRPGEVGPPGPPGPAGEK

900–923

2231.0460

38 38 38 36 36 36

2.7E17 2.7E17 8.6E17 1.6E15 2.8E15 2.8E15

3b

GETGPAGRPGEVGPPGPPGPAGEKe GETGPAGRPGEVGPPGPPGPAGEK GETGPAGRPGEVGPPGPPGPAGEK GETGPAGRPGEVGPPGPPGPAGEK GETGPAGRPGEVGPPGPPGPAGEK GETGPAGRPGEVGPPGPPGPAGEK

900–923

2215.0517

40 38 36 36 34 38

<1.0E20 <1.0E20 <1.0E20 <1.0E20 1.1E15 1.4E15

Three peptides with different amount and positions of Hyp are selected for illustration. The data represent a summary of data received from 4 different collagen isolates (UC). Peptides in bold are found in all measured samples with varied hydroxylation patterns. Both the amount and position of hydroxylation varied. E-value above 103 is considered not significant. a Hydroxylated residues are underlined. Hydroxylation patterns detected with highest probability and frequency are printed in bold. Alternative hydroxylation patterns and their respective E-values are listed separately for comparison. b Relative molecular mass calculated from measured m/z. c Number of fragment ions matched. d Expectation value from program OMSSA. e This pattern was supported by measurements on different peptides. f The ranking of these patterns varied in different measurements.

the UV–CD spectrum was observed until the urea concentration was lowered to 2 M. Below 2 M urea, the characteristic triple helix peak at 225 nm appeared as described previously [9] which was maximal in the absence of urea. We did not observe a timedependent change in the UV–CD spectra taken at concentrations 0–4 M urea over a period of about 2 h.

To characterize the urea-mediated effects further, we obtained UV–vis spectra of the UC in the presence of different concentrations of urea (Fig. 5). In the range 260–300 nm two features could be observed. First, the change in turbidity as the urea concentration was lowered from 8 M to 1 M was not a monotonic function, as would be expected for the formation of increasingly larger non-

Table 3 Differential hydroxylation of proline residues in Col 1a2 chain. Sequencea

Position

Mrb

Ionsc

E-valued

1a

RGSPGEPGSAGPAGPPGLRe RGSPGEPGSAGPAGPPGLR RGSPGEPGSAGPAGPPGLR RGSPGEPGSAGPAGPPGLR

387–405

1747.8601

31 31 28 22

<1.0E20 <1.0E20 3.0E13 4.8E07

1b

GSPGEPGSAGPAGPPGLR GSPGEPGSAGPAGPPGLR GSPGEPGSAGPAGPPGLR

388–405

1607.7545

21 20 17

8.9E12 9.4E11 3.9E06

2a

HGNRGEPGPAGSVGPVGAVGPR HGNRGEPGPAGSVGPVGAVGPRe

981–1002

2040.0249

42 47

1.6E19 2.7E19

2b

HGNRGEPGPAGSVGPVGAVGPR

981–1002

2024.0299

43

1.6E19

Two peptides with different amount and positions of Hyp are selected for illustration. The data represent a summary of data received from 4 different collagen isolates (UC). Peptides in bold are found in all measured samples with varied hydroxylation patterns. Both amount and position of hydroxylation varied. E-value above 103 is considered not significant. (a–e) are defined in Table 2.

X. Xiong et al. / Process Biochemistry 44 (2009) 1200–1212

1207

Fig. 4. UV–CD spectra obtained from the Superose 12-purified collagen polypeptides dissolved in different concentrations of urea. The main panel shows the spectral region accessible in the presence of different concentrations of urea. The characteristic peak at 225 nm corresponding to intact collagen triple helix is indicated by an arrow. The insert shows the complete spectrum obtained from the soluble collagen hydrogel. with the spectra from the main panel superimposed. (B) comparison of the UV–CD spectra of the UC in water obtained here and that of ‘‘pure triple helix’’ obtained using synthetic peptides [9].

specific aggregates, but followed a complex profile (Fig. 5A and B). Strikingly, between 2 M and 4 M urea, where triple helix begins to form, the turbidity dropped to a value below that of the UC in 8 M urea, where only isolated random coil chains are present. This behavior is consistent with the appearance of ordered structures (here triple-helical aggregates), which display smaller effective hydrodynamic radii than the sum of their parts. The second feature observable is the presence of three very weak absorption maxima, with major components at 270 nm, 280 nm and 290 nm (Fig. 5C). To assess the relative intensity changes of the two major maxima we performed second-derivative analysis (which essentially removes the monotonic contribution from turbidity, while ‘‘sharpening’’ the peaks (Fig. 5D)). The urea-profiles of the second derivative values at 265 nm and 287 nm (Fig. 5E) show a striking similarity to that of the turbidity profile (Fig. 5B). The positions of the peaks would correspond well to the peak maxima for tyrosine and its ionized form, tyrosinate. Tyrosine is present in the core

region (5/a1 and 2/a2). At present, we cannot explain why the intensities of the putative tyrosine peaks are so weak. Possibly, the tyrosines exist in a variety of slightly differing environments, which leads to absorption peaks with slightly different maxima, thus ‘‘smearing’’ the intensity over a relatively large range. 3.4. Carboxypeptidase digestion To compare the structural integrity of the UC and AC, dialyzed UC and AC samples were digested with sequencing grade yeast carboxypeptidase Y as described in Section 2. The digested samples were then reduced and analyzed by 6% SDS-PAGE (Fig. 6A). For the enzyme/collagen (E/C) ratio 1/200, digested, reduced samples showed essentially no degradation of both UC and AC for 20 h when analyzed by SDS-PAGE. However, if the E/C weight ratio was increased to 1/50 or 1/25, the AC was digested much faster than the UC as shown in lanes 14–17 (Fig. 6), resulting in complete digestion

Fig. 5. UV-spectral analysis of Superose 12-purified collagen dissolved in various concentrations of urea at 25 8C. (A) UV–vis absorption spectra obtained with different concentrations of urea (indicated). The spectra shown have been obtained after subtraction of the reference which contained the appropriate urea concentration. The concentrations of urea used are indicated. The arrows indicate absorption maximum. (B) Turbidity plotted from the absorptions at 270 nm. (C) Gaussian multicomponent analysis of the UV–vis spectra obtained for collagen dissolved in 2.2 M urea (shown in (A)). (D) Second-derivative analysis of the spectra shown in (A). (E) Urea concentrationdependence of the absorption maxima shown by the arrow in (A).

1208

X. Xiong et al. / Process Biochemistry 44 (2009) 1200–1212

Fig. 6. SDS-PAGE (6%) profile of the carboxypeptidase digestion. The unnumbered lane is yeast carboxypeptidase Y by 61.0 kDa. This band is also be detected in lanes 2–13. Lanes 2 and 8 are the non-digested urea-extracted and acetic acid-extracted collagen respectively. The lanes 3–7 are digestion for 1 min, 5 min, 10 min, 1 h and 20 h; appropriate digestions were performed by AC in the same time series as shown in lanes 9–13; The E/C ratio (w/w) is 1:200 as suggested by manufacturer. Lanes 14–15 show digestions of UC and 16–17 of AC for 20 h with an E/C of 1:50 and 1:25 (w/w). Equal amounts of AC were employed throughout. Thus, the data in lanes 16 and 17 shows that AC was completely digested in this experiment. The band at 61 kDa corresponds to carboxypeptidase.

Fig. 7. SEM images of NIH 3T3 fibroblasts cultivated on the UC and AC coated cover-slips through a time course study up to 8 days. (A and D) The control samples with the NIH 3T3 fibroblasts directly seeded onto the cover-slips; (B and E) NIH 3T3 fibroblasts grown on UC; (C and F) NIH 3T3 fibroblasts grown on AC. SEM images were taken after 4 days (A–C) or 8 days (D–F) of cultivation. The cells are indicated using arrows.

of AC. Equal amounts of AC and UC were employed throughout. The band at 61 kDa corresponds to carboxypeptidase. Therefore, we conclude that UC in this assay is more stable than AC. 3.5. Phenotypic and molecular comparison of NIH 3T3 fibroblasts grown on differently extracted collagen matrices Type I collagen is one of the most commonly used matrices in tissue engineering. Therefore, we determined if the different extraction procedures used in this study (UC vs. AC) have an impact on growth and behavior of mammalian cells. The behavior of NIH 3T3 fibroblast cell cultures when grown on UC and AC were compared by SEM as well as by qRT-PCR. To ensure the accuracy and reliability, both UC and AC used for these experiments are extracted from same batch of tail tendons from freshly killed rats. We could observe significant differences between NIH 3T3 fibroblasts seeded on the different collagen extracts during a time course of up to 8 days (Fig. 7). The NIH 3T3 fibroblasts seeded on AC show abnormal morphology after 4 days as compared to glass grown fibroblasts (Fig. 7C). They were packed closely together with globular form. After 8 days on AC, NIH 3T3 fibroblasts showed mostly deformed morphology without significant amount of lammelipodia (Fig. 7F). In contrast to the ACgrown cells, the UC-grown NIH 3T3 fibroblasts were evenly distributed over the UC surface (Fig. 7B and E) with similar morphology as the control samples directly cultivated on the cover-slips (Fig. 7A and D). The cells grown on UC showed clearly

defined lammelipodia throughout the 8 days and have a spindlelike form which is characteristic for the fibroblasts in a 3D-matrix. To test if the difference in morphology observed in the differently prepared collagen matrices could also be manifested on a molecular level, we performed transcriptional analysis of genes implicated in the regulation of cell adhesion and cell motion, matrix remodeling as well as mechanical stress responses using qRT-PCR. Fak1 was chosen because it was shown to be a central regulator of cell motility and also has a role in mechanical stress response [37]. Fak is also important for sensing the matrix rigidity and for activating the downstream regulatory pathways, which regulate the cell motion and mechanical response. Rac1 and Rhoa have been described as GTPases which are required to generate the force on the ECM required for cell motion. Additionally, Rac1 is also involved in regulation of invasive growth of cells [37]. Both COL1A1 and MMP9 (one of more than 21 known members of metalloproteinases) are known to be involved in matrix remodeling, which is taking place continuously in active tissue [27,37]. TGFb1, TGFb2 and actin are central in response to mechanical stress and a variety of cellular behaviors correlated with traction and ECM synthesis [38–41]. Fak1 can repress expression of Rac1 during non-invasive growth. In our study, upregulated Fak1 and consequently down-regulated Rac1 confirmed the reduced matrix rigidity and mechanical stress level of the fibroblasts in UC if compared to AC (Table 4). A higher stress level in AC-grown NIH 3T3 fibroblasts is also shown by the 23-fold upregulation of TGFb1, when compared to UC-grown NIH 3T3 fibroblasts. Low expression

X. Xiong et al. / Process Biochemistry 44 (2009) 1200–1212

1209

Table 4 Differential gene regulation quantified by real-time PCR. UC Fak1 Rac1 Rho1 MMP9 TGFb1 COL1A1 Actb

10.84 14 1.29 37.64 5 1.18 5.26

7.1 3.84 2.63 42.47 5.09 1.18 11.11

10.64 2 1 2 2 5.27 10

UCmean  deviation

AC

9.53  1.6 6.61  4.9 0.11  1.7 26.04  18.7 0.64  3.8 2.54  1.8 8.79  2.4

3 7.7 3.33 263.16 1.51 25 11.74

ACmean  deviation 1 3.45 8.33 4.76 23.42 1.2 2.57

3.45 3 1.6 100 1 1.44 2.42

2.15  1.88 4.72  1.99 4.42  2.61 122.64  93.68 8.64  9.85 8.25  11.16 5.58  4.11

Values represent mean values from three different experiments (7 days cultivation) using different batches of collagen. Values are compared between UC/AC and glass surface.

of MMP9 and CoL1A1 in AC indicates low matrix remodeling activity when compared to the expression level of both genes in UC grown fibroblasts. Thus, the results of our qRT-PCR analysis confirmed our microscopic data. The results suggest that UC may be a preferred matrix for growth of fibroblasts than AC and therefore could serve as an improved matrix component for tissue engineering. 4. Discussion We have established a fast two-step urea-extraction procedure to isolate and purify Type I collagen from rat tail tendon. Although the extraction of collagen Type I using urea was reported more than 30 years ago [18,42,43], this method has been limited in its application maybe due to reports that only low quality collagen can be obtained with this method. In our study, we found that the collagen isolated using 9 M ultrapure urea is of high quality and shows a reversible aggregation behavior after purification using Superose 12 chromatography (in 8 M urea) (Fig. 1). The extracted collagen was stable in 8 M urea solution over the entire extraction and purification procedure as observed by SDS-PAGE. In contrast, in an earlier study, urea-extracted collagen was reported to be of poor quality [18]. However, the low purity of commercially available urea at that time may have biased these results [44]. Unlike UC, AC could not be dissolved in other non-acidic solutions including urea. Interestingly, we observed a complex molecular weight pattern for UC after Superose 12 chromatography. This pattern cannot be due to cross-linked high molecular weight aggregates of collagen or contaminating proteins for several reasons. First of all, the rat tail tendon derived from 3-month-old rats is known to have very few lysyl-cross-links [45]. Secondly, only a single mass of about 96 kDa was observed in the MALDI-TOF MS spectra for each of the fractions analyzed (Fig. 2A). Sinapic acid, used as a MALDI-matrix, had no effect on the mass spectra and peak intensity during incubation for up to 60 h. This is in agreement with data from Eriksen et al. [46] which showed no effect of MALDI-matrices on the degree of cross-linking of model collagen telopeptides. Third, further analysis of UC samples showed two sub-peaks (with mass of 94 and 96 kDa) indicating the presence of both a1 and a2 chains. No other protein was detected using MS excluding the presence of significant amounts of contaminants in all fractions tested (Fig. 2B). In addition, the amino acid composition analysis of UC samples showed good agreement with the theoretical values (SwissProt), particularly with respect to the amounts of Gly and Pro/Hyp (Fig. 3B and Supplementary Table S1). The difference between the theoretical and measured values was within the expected error range due to degradation of individual amino acids and other known effects during the hydrolysis [44,47]. The MALDI spectra of all fractions obtained after Superose 12 gel-filtration also confirmed the homogeneity and purity of the UC (Fig. 2B). The collagen-recovery using the UCprocedure reported here is about 65%. The low absorbance at 280 nm may have resulted from the low amount of aromatic amino

acids especially the lack of tryptophan. In contrast to UC, AC preparations could not be further purified by Superose 12 chromatography due to precipitation of the collagen. In addition, other proteins [48], e.g. albumin, were also identified in AC in our ESI-MS/MS experiments. Thus, AC extracted according to a standard procedure used in industrial production shows significant differences to UC. Fourthly, re-chromatography of the individual UC fractions showed that the observed aggregation is reversible, since pooled fractions of each sub-peak resulted in similar chromatographic and SDS-PAGE profile, showing both lower and higher molecular weight aggregates as initially observed. The SDS-PAGE profiles also indicate that the Superose 12 column had removed other components which probably enhance collagen gelling, such as proteins interacting with collagen [49–51], indicated by removal of the ‘‘smear’’ observable for SDS-PAGE profile of the unpurified UC (compare Fig. 1, lane 1 to lanes 2–16). The removal of these unknown components may lead to the high solubility of collagen in water observed. Although ultra-pure urea was used in the extraction procedure we observed that 25% of all Lys residues were carbamylated, a typical reaction occurring during long term incubation of proteins in urea. This carbamylation could be a reason for the enhanced solubility and reversibility of the reversible aggregation pattern observed. Brodsky et al. [52], suggested that in the absence of propeptides, a change in charge of the N- and C-termini (e.g. binding of glycoprotein) of the a1 and a2 chains may result in the formation of triple helices in vivo. It is likely that carbamylation of Lys residues changes the local charge, thereby promoting a stable folding of the collagen a chains to a triple helix-like structure in purified UC. This observation is consistent with the results from Jaisson et al. [53–56]. 4.1. Detailed sequence analysis of UC using MS The ESI-MS approaches using fractions from all the 3 sub-peaks, successfully covered 96% and 88% of the sequences of collagen Type I a1 and a2 respectively (Fig. 3A and Supplementary Fig. S2). To our knowledge, this is the first time that such a high coverage of the protein sequences of Type I collagen could be obtained. This confirmed the high structural integrity and purity of the UC prepared in this study. The ESI-MS/MS analysis also showed that no other proteins could be detected in these samples in each peak. The high molecular weight bands (all individual bands were analyzed using ESI-MS/MS) (Fig. 1B) were also Type I collagen containing no other detectable high molecular weight contamination as described for AC [44]. Hydroxyproline (Hyp) is very important for the stability of collagen [52], and is thus important for the assembly, folding, stability and biological functions of the collagen. Recent work has proposed that Hyp is required for the triple-helix self-assembly [57]. However, most of the studies on collagen assembly are based on model peptides which consist of small numbers of [Gly-X-Y] repeats, which are more conducive for structural analysis of

1210

X. Xiong et al. / Process Biochemistry 44 (2009) 1200–1212

collagen using NMR and X-ray crystallography. The high sequence coverage of our MS/MS data allow us not only to identify the hydroxylation sites but also to define the sequence position of Hyp within the entire protein, hence facilitating further detailed studies of the hydroxylation pattern in native collagen. We could show that 92–120 proline residues from 236 Pro in a1 chain and 70–90 Pro from 201 Pro in a2 chain were detected as Hyp in the UC samples isolated. A very interesting result from the MS/MS data is the differential hydroxylation of proline residues. Some peptides with the same primary sequence showed distinct hydroxylation patterns as shown in Tables 2 and 3. This indicates that the Type I collagen contains not only [a1]2a2 but differentially modified a1 and a2 chains, which may result in different roles during assembly, folding and biological function of collagen. Moreover, in some peptides, the Hyps were found at the X-position of the [Gly-X-Y] repeats which has not been observed previously. We observed these Hyp-patterns both in the UC and AC samples investigated indicating that additional prolyl-hydroxylases with different specificities may exist in mammalian genome or that the known prolyl-hydroxylases might have altered specificities under in vivo conditions. Therefore, we hypothesize that the collagen stability/ flexibility in different tissues may be regulated at least partially by these differential hydroxylation patterns. This would potentiate the complexity of collagen as a scaffold for interaction with other components. We observed a significant mass difference between UC (96 kDa) and AC (93.5 kDa) of 2.5 kDa. The mass difference between AC and UC cannot be attributed entirely to the carbamylation observed in UC during extraction, as the total carbamylation detected in UC is lower than 1 kDa. Additionally, for UC we observed two mass peaks (96 kDa and 94.5 kDa) most likely corresponding to the a1 and a2 chains respectively. For AC only one mass peak could be resolved. As indicated in the SwissProt data, the a1 and a2 chains should have masses of about 94.8 kDa and 91.7 kDa respectively. If we calculate the hydroxylation of proline residues detected in the ESIMS/MS data, we obtain a 2.1 kDa increase of the mass, resulting in final masses of 96.9 kDa and 93.8 kDa for the a1 and a2 chains respectively. This accounts well for the masses observed experimentally for UC. Therefore, the two peaks detected in UC samples most likely represent the hydroxylated a1 and a2 chains. Previous studies reported the degradation of collagen during acidic extraction [18,42,44], which could be an explanation of the lowered mass of AC. In agreement with this observation, sequence coverage of AC was not complete, particularly at both the N- and Ctermini, which indicates that the terminal regions are more accessible to non-specific acid hydrolysis. This is consistent with the lower stability of the AC in the carboxypeptidase digestion. In addition, it is well known that collagen extracted using either neutral salt or acidic solution must be either lyophilized or stored cold in acetic acid to avoid degradation [16]. An exact determination of the reason for the mass difference observed between UC and AC, however, was not possible from the data obtained by MS. 4.2. Spectroscopic analysis of UC After dialysis against water and lyophilization the ureaextracted collagen could be redissolved in water to yield a clear sol–gel type solution. The results obtained by UV–CD spectroscopy strongly suggest that the urea-extracted collagen dissolved in water exhibits a high proportion of triple helical structure. The complex behavior of the absorption maxima in UV–CD- and UV–vis-spectra during reversible aggregation of UC collagen suggests strongly that the assembly process itself is complex, probably proceeding through several conformational/aggregational states. Most interestingly, we observed this self-assembly process in urea-solution of different concentrations in the absence

of detectable N- or C-terminal propeptides, which have been shown to be important for the triple helix formation [58]. To our knowledge, the reversible aggregation of a collagen fraction consisting only of core region in urea is shown here for the first time. Characterization of the macromolecular structure of UC after extensive dialysis against water and lyophilization using SEM showed large structures (reassembled fibrils) (diameter > 2 mm) in the UC samples (Supplementary Fig. S5B). Although the AC was also dialyzed and lyophilized in parallel to UC, however, we obtained only a collagen-film without any detectable fibrillar structure (Supplementary Fig. S5C). In contrast to native rat tail tendons, the banding pattern (D-period) could not be observed in UC indicating only partial renaturation of complete collagen fibrils. This inability of fibrillogenesis may be a result of carbamylation, which results in a change of the charge in both ends of the collagen chains and thus may hinder the cross-linking and fibril formation of UC. This effect could contribute to the high water solubility of UC and the observed growth behavior of the fibroblasts on UC. 4.3. Structural integrity Structural integrity of UC and AC was compared by enzymatic digestion using carboxypeptidase Y. UC shows a high resistance against yeast carboxypeptidase Y (Fig. 6), indicating that a stable structure is formed after dialysis, which is in agreement with a triple helical structure as indicated by UV–CD spectroscopy [59,60]. The dialyzed AC samples were almost completely digested by carboxypeptidase (Fig. 6, lanes 16 and 17), indicating a less compact structure. The carbamylation of Lys residues in UC may also contribute to the resistance to carboxypeptidase Y either by reducing protease activity or by promoting the formation of a native-like triple helical structure as discussed above. 4.4. Compatibility with tissue culture Type I collagen is a major component of matrix for tissue engineering. As reported previously, the cell–matrix interaction can control cellular phenotypes [61]. Therefore, we tested, if we could observe differences between AC and UC matrices with respect to morphology, adhesion and motility of mammalian cells on these distinct surfaces. In this report, we show that the NIH 3T3 fibroblasts could attach and extend themselves easily on UCcoated cover-slips similar to glass cover-slips (Fig. 7A, B, D and E). Growth on the AC surface, however, resulted in closely-packed cells (Fig. 7C and E). Moreover, the cell morphology between the cells grown on UC and AC differed strongly. The cells on UC as well as on glass slides show lammelipodia formation (Fig. 7A, B, D and E) whereas the cells grown on AC were rounded and lammelipodia could be barely observed (Fig. 7C, F). UC is not as solid as AC. This might be a reason that the cells grown on UC show 3D-like morphology and higher motility. In addition, Jaisson et al. have reported that the carbamylation of Type I collagen can alter the cellular function [53–56]. How the carbamylation detected in UC influences the growth of fibroblasts, however, has not been studied previously and should be investigated in future studies more in detail. The transition from initially round (non-motile) to polarized and spindle-shaped implicates the development of a tensile force between the cell edges. Many studies have shown that changes in cell shape require cytoskeletal action and are accompanied by the redistribution of cell surface receptors including integrins, signal induction, cell spreading and/or contraction, and the induction of gene expression manifested by both degradation and synthesis of extracellular macromolecules [62–66]. To verify if the morphological differences observed are reflected at the molecular level, we tested if genes involved in regulation of cell motility, mechanical stress response and matrix remodeling

X. Xiong et al. / Process Biochemistry 44 (2009) 1200–1212

are regulated differentially in UC- and AC-grown NIH 3T3 fibroblasts. Real-time PCR quantification of several genes involved in regulation of cell motion, adhesion and matrix remodeling showed significant differences of NIH 3T3 fibroblasts grown on UC as compared to AC. It is well known that cells have decreased motility if they are cultivated in a rigid ECM, which is sensed by Fak1. The more active Fak1, the faster the cell can migrate. Rho generates the force on the ECM as well as Rac. It has been shown that in rapidly migrating cells the expression of FAK1 is strongly up-regulated [37]. Interestingly, Fak1 was strongly up-regulated in cells grown on UC if compared to AC, consequently Rac1 is downregulated since Fak1 repress Rac1 in non-invasive growth [37]. The Rac pathway is required for invasive growth of the cells into ECM through the PI3K pathway. Matrix remodeling, particularly collagen remodeling can also influence cellular behavior such as motion and proliferation. The active matrix-remodeling indicated by upregulation of MMP9 and COL1A1 in cells grown on UC confirmed the higher activity of NIH 3T3 fibroblasts on UC than on AC. The upregulation of MMP9 has also been observed by Garnotel et al. [53]. They indicated that differences in gene expression may result from the carbamylation of Type I collagen in UC matrix. The absolute expression ratios did vary in different experiments, however, induction or repression of the respective gene is reproducible. In summary, we have provided a new method for rapid, high yield purification of Type I collagen both for molecular analysis and application in tissue engineering. We could show high sequence coverage using MS/MS techniques resulting in the discovery of differential hydroxylation patterns of proline residues in the [GlyX-Y] repeats. This includes Hyp at the X-position which was not observed previously. In addition, our data indicate that ureaextracted collagen may have advantages to be applied as scaffold or matrix for tissue engineering and cell cultures in general. Acknowledgements We would like to thank Mrs. M. Riedl and Dr. H. Weber for the technical support. We also thank Dr. W. Haehnel for kindly providing mass spectrometry facilities. We thank Ms. C. Autenrieth for performing the Gaussian analysis. Especially, we want to thank Prof. A. Veis for fruitful discussions. This work was supported by a fellowship of the Peter und Traudl Engelhorn Foundation to X.X. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.procbio.2009.06.010. References [1] Kadler KE, Baldock C, Bella J, Boot-Handford RP. Collagens at a glance. J Cell Sci 2007;120:1955–8. [2] Rosso F, Giordano A, Barbarisi M, Barbarisi A. From cell-ECM interactions to tissue engineering. J Cell Physiol 2004;199:174–80. [3] Berglund JD, Mohseni MM, Nerem RM, Sambanis A. A biological hybrid model for collagen-based tissue engineered vascular constructs. Biomaterials 2003;24:1241–54. [4] Purna SK, Mary B. Collagen based dressings — a review. Burns 2000;26:54–62. [5] Wallace DJ, Rosenblatt J. Collagen gel systems for sustained delivery and tissue engineering. Adv Drug Deliv Rev 2003;55:1631–49. [6] Dieterich C, Schandar M, Noll M, Johannes FJ, Brunner H, Graeve T, et al. In vitro reconstructed human epithelia reveal contributions of Candida albicans EFG1 and CPH1 to adhesion and invasion. Microbiology 2002;148:497–506. [7] Becker U, Timpl R, Kuhn K. Carboxyterminal antigenic determinants of collagen from calf skin. Localization within discrete regions of the nonhelical sequence. Eur J Biochem 1972;28:221–31. [8] Miller EJ, Gay S, Leon WC. The collagens: an overview and update. Methods in enzymology, vol 144. Academic Press; 1987. p. 3. [9] Kotch FW, Raines RT. Self-assembly of synthetic collagen triple helices. Proc Natl Acad Sci USA 2006;103:3028–33.

1211

[10] Kar K, Amin P, Bryan MA, Persikov AV, Mohs A, Wang Y-H, et al. Self-association of collagen triple helic peptides into higher order structures. J Biol Chem 2006;281:33283–90. [11] Kadler KE, Hojima Y, Prockop DJ. Assembly of type I collagen fibrils de novo, between 37 and 41 8C the process is limited by micro-unfolding of monomers. J Biol Chem 1988;263:10517–23. [12] Bornstein P. The NH(2)-terminal propeptides of fibrillar collagens: highly conserved domains with poorly understood functions. Matrix Biol 2002;21:217–26. [13] Olsen BR, Guzman, Norberto A, Engel J, Condit C, Aase S. Purification and characterization of a peptide from the carboxy-terminal region of chick tendon procollagen type I. Biochemistry 1977;16:3030–6. [14] Ottani V, Raspanti M, Ruggeri A. Collagen structure and functional implications. Micron 2001;32:251–60. [15] Brodsky B, Persikov AV. Molecular structure of the collagen triple helix. Adv Protein Chem 2005;70:301–39. [16] Kessler A, Rosen H, Levenson SM. Chromatographic fractionation of acetic acid-solubilized rat tail tendon collagen. J Biol Chem 1960;235:989–94. [17] Rajan N, Habermehl J, Cote M-F, Doillon CJ, Mantovani D. Preparation of readyto-use, storable and reconstituted type I collagen from rat tail tendon for tissue engineering applications. Nat Protocols 2007;1:2753–8. [18] Adelmann BC. The structural basis of cell-mediated immunological reactions of collagen. Reactivity of separated-chains of calf and rat collagen in cutaneous delayed hypersensitivity reactions. Immunology 1972;23:739–48. [19] Becker U, Timpl R. NH2-terminal extensions on skin collagen from sheep with a genetic defect in conversion of procollagen into collagen. Biochemistry 1976;15:2853–62. [20] Ruszczak Z, Friess W. Collagen as a carrier for on-site delivery of antibacterial drugs. Adv Drug Deliv Rev 2003;55:1679–98. [21] Weinberg CB, Bell E. A blood vessel model constructed from collagen and cultured vascular cells. Science 1986;231:397–400. [22] Ratner BD, Bryant SJ. Biomaterials: where we have been and where we are going. Annu Rev Biomed Eng 2004;6:41–75. [23] Goissis G, Marcantonio Jr E, Marcantonio RA, Lia RC, Cancian DC, de Carvalho WM. Biocompatibility studies of anionic collagen membranes with different degree of glutaraldehyde cross-linking. Biomaterials 1999;20:27–34. [24] Chvapil M, Kronenthal L, Van Winkle Jr W. Medical and surgical applications of collagen. Int Rev Connect Tissue Res 1973;6:1–61. [25] Cukierman E, Pankov R, Yamada KM. Cell interactions with three-dimensional matrices. Curr Opin Cell Biol 2002;14:633–9. [26] Di Lullo GA, Sweeney SM, Korkko J, Ala-Kokko L, San Antonio JD. Mapping the ligand-binding sites and disease-associated mutations on the most abundant protein in the human, type I collagen. J Biol Chem 2002;277:4223–31. [27] Cowin SC. Tissue growth and remodeling. Annu Rev Biomed Eng 2004;6:77–107. [28] Laemmli UK. Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 1970;227:680–5. [29] Geer LY, Markey SP, Kowalak JA, Wagner L, Xu M, Maynard DM, et al. Open mass spectrometry search algorithm. J Proteome Res 2004;3:958–64. [30] Acil Y, Mobasseri AE, Warnke PH, Terheyden H, Wiltfang J, Springer I. Detection of mature collagen in human dental enamel. Calcif Tissue Int 2005;76:121–6. [31] Lester DE. Computerized resolution of overlapping bands in UV spectra using Gaussian profile approximations. Anal Biochem 1970;36:253–67. [32] Habermehl J, Skopinska J, Boccafoschi F, Sionkowska A, Kaczmarek H, Laroche G, et al. Preparation of ready-to-use, stockable and reconstituted collagen. Macromol Biosci 2005;5:821–8. [33] Deyl Z, Miksik I, Eckhardt A. Preparative procedures and purity assessment of collagen proteins. J Chromatogr B Analyt Technol Biomed Life Sci 2003;790:245– 75. [34] Atkins P, de Paula J. Atkins’ physical chemistry, 6th ed., Oxford: Oxford University Press; 2002. p. 721–5. [35] Brinker CJ, Scherer GW. Sol–gel science: the physics and chemistry of sol–gel processing. Boston: Academic Press; 1990. p. 1–13. [36] Ma B, Zhang K, Hendrie C, Liang C, Li M, Doherty-Kirby A, et al. PEAKS: powerful software for peptide de novo sequencing by tandem mass spectrometry. Rapid Commun Mass Spectrom 2003;17:2337–42. [37] Li S, Guan JL, Chien S. Biochemistry and biomechanics of cell motility. Annu Rev Biomed Eng 2005;7:105–50. [38] Schild C, Trueb B. Three members of the connective tissue growth factor family CCN are differentially regulated by mechanical stress. Biochim Biophys Acta 2004;1691:33–40. [39] Wang Z, Fong KD, Phan TT, Lim IJ, Longaker MT, Yang GP. Increased transcriptional response to mechanical strain in keloid fibroblasts due to increased focal adhesion complex formation. J Cell Physiol 2006;206:510–7. [40] Kobayashi T, Liu X, Kim HJ, Kohyama T, Wen FQ, Abe S, et al. TGF-beta1 and serum both stimulate contraction but differentially affect apoptosis in 3D collagen gels. Respir Res 2005;6:141. [41] Fong KD, Trindade MC, Wang Z, Nacamuli RP, Pham H, Fang TD, et al. Microarray analysis of mechanical shear effects on flexor tendon cells. Plast Reconstr Surg 2005;116:1393–404. discussion 1405–1396. [42] Becker U, Timpl R. Cyanogen bromide peptides of the rabbit collagen a1-chain. FEBS Lett 1972;27:85–8. [43] Rauterberg J, Fietzek P, Rexrodt F, Becker U, Stark M, Kuhn K. The amino acid sequence of the carboxyterminal nonhelical cross link region of the alpha 1 chain of calf skin collagen. FEBS Lett 1972;21:75–9. [44] Chandrakasan G, Torchia DA, Piez KA. Preparation of intact monomeric collagen from rat tail tendon and skin and the structure of the nonhelical ends in solution. J Biol Chem 1976;251:6062–7.

1212

X. Xiong et al. / Process Biochemistry 44 (2009) 1200–1212

[45] Eyre D, Leon WC. Collagen cross-linking amino acids. In: Methods in enzymology. Paris: Academic Press; 1987. pp. 115–139. [46] Eriksen HA, Sharp CA, Robins SP, Sassi ML, Risteli L, Risteli J. Differently crosslinked and uncross-linked carboxy-terminal telopeptides of type I collagen in human mineralised bone. Bone 2004;34:720–7. [47] Piez KA, Gross J. The amino acid composition of some fish collagens: the relation between composition and structure. J Biol Chem 1960;235:995–8. [48] Miller EJ, Gay S, Leon WC. Collagen: an overview. In: Methods in enzymology. Paris: Academic Press; 1982. p. 3–32. [49] Bann JG, Bachinger HP, Peyton DH. Role of carbohydrate in stabilizing the triple-helix in a model for a deep-sea hydrothermal vent worm collagen. Biochemistry 2003;42:4042–8. [50] Mizuno K, Hayashi T, Peyton DH, Bachinger HP. Hydroxylation-induced stabilization of the collagen triple helix: acetyl-(glycyl-4(R)-hydroxyprolyl-4(R)-hydroxyprolyl)10-NH2 forms a highly stable triple helix. J Biol Chem 2004;279:38072–8. [51] Parodi AJ. Protein glucosylation and its role in protein folding. Annu Rev Biochem 2000;69:69–93. [52] Brodsky B, Thiagarajan G, Madhan B, Kar K. Triple-helical peptides: An approach to collagen conformation, stability, and self-association. Biopolymers 2008;89:345–53. [53] Garnotel R, Sabbah N, Jaisson S, Gillery P. Enhanced activation of and increased production of matrix metalloproteinase-9 by human blood monocytes upon adhering to carbamylated collagen. FEBS Lett 2004;563:13–6. [54] Jaisson S, Larreta-Garde V, Bellon G, Hornebeck W, Garnotel R, Gillery P. Carbamylation differentially alters type I collagen sensitivity to various collagenases. Matrix Biol 2007;26:190–6. [55] Jaisson S, Lorimier S, Ricard-Blum S, Sockalingum GD, Delevallee-Forte C, Kegelaer G, et al. Impact of carbamylation on type I collagen conformational structure and its ability to activate human polymorphonuclear neutrophils. Chem Biol 2006;13:149–59.

[56] Jaisson S, Sartelet H, Perreau C, Blanchevoye C, Garnotel R, Gillery P. Involvement of lysine 1047 in type I collagen-mediated activation of polymorphonuclear neutrophils. FEBS J 2008;275:3226–35. [57] Mohs A, Silva T, Yoshida T, Amin R, Lukomski S, Inouye M, et al. Mechanism of stabilization of a bacterial collagen triple helix in the absence of hydroxyproline. J Biol Chem 2007;282:29757–65. [58] Khoshnoodi J, Cartailler JP, Alvares K, Veis A, Hudson BG. Molecular recognition in the assembly of collagens: terminal noncollagenous domains are key recognition modules in the formation of triple helical protomers. J Biol Chem 2006;281:38117–21. [59] Bruckner P, Prockop DJ. Proteolytic enzymes as probes for the triple-helical conformation of procollagen. Anal Biochem 1981;110:360–8. [60] Layman DL, McGoodwin EB, Martin GR. The nature of the collagen synthesized by cultured human fibroblasts. Proc Natl Acad Sci USA 1971;68:454–8. [61] Friedl P, Brocker EB. The biology of cell locomotion within three-dimensional extracellular matrix. Cell Mol Life Sci 2000;57:41–64. [62] Friedl P, Zanker KS, Brocker EB. Cell migration strategies in 3-D extracellular matrix: differences in morphology, cell matrix interactions, and integrin function. Microsc Res Tech 1998;43:369–78. [63] Heino J. Biology of tumor cell invasion: interplay of cell adhesion and matrix degradation. Int J Cancer 1996;65:717–22. [64] Aggeler J, Frisch SM, Werb Z. Changes in cell shape correlate with collagenase gene expression in rabbit synovial fibroblasts. J Cell Biol 1984;98: 1662–71. [65] Akiyama SK, Olden K, Yamada KM. Fibronectin and integrins in invasion and metastasis. Cancer Metastasis Rev 1995;14:173–89. [66] Langholz O, Rockel D, Mauch C, Kozlowska E, Bank I, Krieg T, et al. Collagen and collagenase gene expression in three-dimensional collagen lattices are differentially regulated by alpha 1 beta 1 and alpha 2 beta 1 integrins. J Cell Biol 1995;131:1903–15.