Haploid Saccharomyces cerevisiae BY4741-ΔTRP1 (strain BY4741 with a TRP1 deletion: MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 Δtrp1:: KanMX) 18 was used to express the modular collagen and the human prolyl-4-hydroxylase. E. coli strain XL1-Blue (Stratagene) was used for plasmid maintenance and amplification. Human prolyl-4-hydroxylase subunit genes p4Ha and p4Hb were cloned into a low copy CEN/ARS vector and co-transformed with the 2 μ modular collagen plasmid into BY4741-ΔTRP1 as described previously 19. The p4Ha gene was cloned under the GAL10 promoter and the p4Hb gene with a yeast invertase (SUC2) leader sequence was cloned under the GAL1 promoter on the same CEN/ARS plasmid. The modular collagen gene was under the control of the S. cerevisiae GAL1 promoter 2. To test the effect of balancing the collagen-to-hydroxylase gene copy ratio, the collagen gene was placed on a CEN/ARS plasmid. The GAL1p-MCOL gene cassette was excised using HindIII from YEpMCOL 2, inserted into YCplac22 20 to create YCpMCOL-trp, a low-copy CEN/ARS vector. This vector was co-transformed with the CEN/ARS hydroxylase vector into BY4741-ΔTRP1. As a negative control strain (no hydroxylase present), YEpMCOL was transformed into BY4741-ΔTRP1 alone without the hydroxylase plasmid.

Strains carrying the dual plasmids were inoculated into selective medium and cultured as described in Chan et al. 2. Cells were lysed and their background protein was digested with pepsin. Recombinant collagen was purified from the digested protein by cationic exchange chromatography as described in Chan et al. 2.

Samples of dried bovine collagen (5 to 12 μg; Millipore) and FPLC-purified recombinant collagen samples (40 to 80 μg) were hydrolyzed in 6 N hydrochloric acid (Pierce) vapor at 150°C for 90 minutes under vacuum 15. The hydrolysate was resuspended with acetonitrile/water (50:50 v/v) containing 0.5 μg/ml of the IS and injected onto the LC-MS system. For comparison, AAA was performed by the Biopolymer Lab at UCLA (http://www.mbi.ucla.edu/services/biopolymer-laboratory). For AAA, protein samples were hydrolyzed in 6 N (vapor-phase) HCl for 22 hours at 110°C. The free amino acids were then derivatized with 6-aminoquinoly-N-hydroxysuccinimidyl carbamate (AQC, Waters), a fluorescent amino-reactive probe, and the products were fractionated and quantified by reversed-phase HPLC.

To investigate the feasibility of isolating collagen from SDS-PAGE for analysis, SDS-PAGE loading buffer (6×) was added to collagen samples (totalling 320 μl), heated at 95°C for 5 minutes, and run on a 7% SDS-PAGE gel. The gel was washed with water, stained in 0.1% Coomassie Blue R-250 (Fisher) dissolved in 10% acetic acid/50% methanol/40% water, and then destained in 10% acetic acid/50% methanol/40% water to determine band locations. The collagen bands were excised, placed in a glass vial, further destained by washing three times in 10% acetic acid/50% methanol/40% water, and stored at 4°C overnight. To extract the collagen, acetonitrile (Fisher, Optima LC-MS grade, 100 μl) was added to the vial that was then sonicated in a water bath for 20 minutes at 40 kHz. The acetonitrile was recovered and a second extraction with 50 μl fresh acetonitrile was performed. The combined acetonitrile extracts were dried under nitrogen in a LC-MS vial and then subjected to hydrolysis as described for FPLC-purified samples. The products were resuspended with 0.1 ml acetonitrile:water (50:50 v/v) containing 0.5 μg/ml IS and analyzed by LC-MS.

The HYP present in collagen is the product of in vivo hydroxylation of PRO residues 21, and we defined the percent hydroxylation of proline to be (moles of HYP)/(moles of HYP + moles of PRO) × 100%. The percent hydroxylation was calculated using concentrations obtained from the LC-MS data. Final values presented in Table 1 are the result of averaging triplicate measurements from 2–3 independent experiments for LC-MS, and measurements of triplicate experiments for AAA. The reported percent hydroxylation values obtained by AAA have the background chemical noise (0.18%) subtracted; this was obtained from the yeast strain containing the collagen gene, but without hydroxylase genes (negative control). The background signal observed in the LC-MS for this negative control was less than 1% relative to readings of any of the HYP standards (Figure 1) and was therefore considered insignificant.

* "N.D." denotes no significant signal of hydroxyproline above background noise.

^§Only 1 AAA result available.

The peaks for proline (PRO), hydroxyproline (HYP), and the internal standard (IS) were identified from injecting pure samples in the LC-MS system. The choice of the internal standard was based on stability, similarity of chemical nature, molecular mass (when compared with PRO and HYP), and commercial availability. Elution times were 1.6 minutes for HYP, 1.65 minutes for PRO, and 2.25 minutes for the IS, glycyl phenylalanine. These elution times were expected for non-derivatized amino acids on the reversed-phase column. The peak area ratios between the fixed amount of internal standard and the two amino acids, proline and hydroxyproline, yielded calibration curves that increased linearly with the amino acid concentrations (Figure 1). Typical linear regression equations were y = 0.144x + 0.006 for PRO and y = 0.288x + 0.019 for HYP, with the coefficients of regression (R²) greater than 0.994. Therefore, for unknown protein hydrolysates, the peak area ratios for the two amino acids against the internal standard can be used to determine the amino acid relative concentrations.

To evaluate our LC-MS method, we examined percent hydroxylation of commercial bovine collagen III and recombinant collagen from two S. cerevisiae strains engineered to produce prolyl-4-hydroxylase. From prior amino acid analysis (AAA), we expected that the percent hydroxylation for the recombinant systems would yield intermediate values that are less than those of native (bovine) collagen and therefore enable us to probe the full range of possible hydroxylation values. Figure 2 shows reconstructed ion chromatographs (RIC) of representative collagen samples. Based on the mass and elution times, the PRO, HYP, and IS peaks were identified. The integrated areas under the PRO and HYP peaks were divided by the integrated area of the IS and their relative concentrations were obtained using the respective standard curves. Percentage proline hydroxylation was then determined as described in Methods.

The percentage proline hydroxylation values are summarized in Table 1. For native bovine collagen III (positive control), the percent hydroxylation of proline was determined by our LC-MS assay to be 38 ± 1.1% (n = 3 separate samples, with each sample injected in triplicate). The percentage hydroxylation of bovine collagen was also determined by amino acid analysis (AAA), which showed 37 ± 2.7% hydroxylation. These values are consistent with a previously-reported value of 41% hydroxylation for bovine collagen as determined by AAA 22. For the negative control, we placed a collagen gene on a 2-micron plasmid in yeast that did not contain the prolyl hydroxylase genes. As expected, neither the LC-MS nor AAA methods yielded hydroxylation signals beyond background chemical noise.

To examine the ability of the LC-MS assay to determine intermediate hydroxylation levels, we evaluated collagen produced from a 2-micron vector (high copy number) and a CEN/ARS (low copy number) vector. In both cases, the hydroxylase enzymes were co-expressed with collagen and carried on a CEN/ARS vector. For these recombinant samples, our LC-MS method obtained percent hydroxylation of 5.8 ± 0.5% for collagen on the 2-micron (high copy number) vector and 12 ± 1.8% for the CEN/ARS (low copy number) plasmid. In comparison, AAA of these samples gave 5.2 ± 1.3% and 12% hydroxylation, respectively. Collectively, our data shows that for determining the percentage of hydroxylated proline, the LC-MS method yields values equivalent to the conventional strategy of AAA.

Furthermore, the precision for the hydroxylation measurements using our LC-MS method for bovine, recombinant (2-micron), and recombinant (CEN/ARS) collagen samples were 2.9%, 7.9%, and 15% relative standard deviation (RSD), respectively. In contrast, using AAA, the RSD for bovine collagen and recombinant collagen (2-micron) were 7.3% and 23%, respectively. This demonstrates that the LC-MS method is also at least as precise, if not more so, as conventional AAA.

Many AAA systems have a sensitivity of 1 pmole 15, with a recommended minimum of 50 pmoles for quantitative analysis. Our experience with AAA has shown that this minimum level of 50 pmoles is especially critical for the quantification of HYP. Below this level, the elution time of HYP in AAA is very close to that of the hydrolysis by-product, 6-aminoquinoline, resulting in overlapping baselines in the chromatogram. In our LC-MS assay, the lower limit of quantification for the pure amino acids is approximately 15 pmole on column for both PRO and HYP. In addition, because reconstructed ion chromatographs are used, there is no problem of an overlap with 6-aminoquinoline. Therefore, the amount of samples needed for both our LC-MS assay and AAA are comparable in magnitude, with the LC-MS assay requiring slightly less sample.

The recombinant human collagen variants presented in Table 1 were purified by fast protein liquid chromatography (FPLC), as described in Methods. However, FPLC can be time-consuming and not practical for larger-scale screens of protein variants, so we investigated whether collagen that was extracted from SDS-PAGE could yield reasonable protein hydroxylation values using the LC-MS assay. Success in quantifying HYP from these extracts would enable analysis of samples without development of lengthy purification protocols.

Although purified by the commercial source, the bovine collagen still shows impurity bands when run on SDS-PAGE gels. The percent hydroxylation for bovine collagen III extracted from SDS-PAGE gave 36 ± 3.4%, which was close to the values measured from the as-received samples (38 ± 1.1%). However, the RSD (an indicator of repeatability) of bovine collagen extracted from SDS-PAGE was 9.4%, which is higher than the 2.9% RSD of as-received collagen, suggesting that the additional steps and reagents required in SDS-PAGE extraction may decrease the precision of our method.

For recombinant human collagen extracted from SDS-PAGE gels, we found that LC-MS could indeed detect the significantly lower levels of hydroxylation present (relative to fully hydroxylated bovine collagen); however, these results were not as accurate or precise as for FPLC-purified samples. This demonstrates that extraction from SDS-PAGE could potentially be used to screen protein candidates to identify desired levels of hydroxylation, but chromatography-based purification may be necessary for in-depth analysis.

We have assayed the proline hydroxylation levels of recombinant collagen produced in two dual-plasmid yeast strains. The first strain overexpressed the collagen using a 2 μ plasmid and produced a 5.8% hydroxylated protein. The second strain expressed collagen from a stable, low copy CEN/ARS plasmid, and in this case the hydroxylation increased to 12%. In both cases, the two hydroxylase genes were carried on a second CEN/ARS plasmid. Therefore, the two strains have different collagen to hydroxylase gene ratios, and this is reflected in the percent hydroxylation of the collagen produced. When the copy number of the collagen gene was high relative to the hydroxylase genes, we obtained a lower level of hydroxylation.

Native collagen has a significantly greater proline hydroxylation level (38% for bovine collagen III) than our systems reported here. Based on the findings from the two dual plasmid systems, the percentage hydroxylation should increase if the hydroxylase and collagen expression levels are appropriately balanced. We expect that further increasing the number of p4H genes or their expression levels should increase hydroxylation. In light of this, our current work includes integrating the two hydroxylase subunits and the collagen gene cassettes into the yeast genome to control tightly the collagen-to-prolyl hydroxylase gene copy ratio. Our facile LC-MS assay will be very helpful in determining the percentage hydroxylation of these integrated strains and identifying the construct with hydroxylation levels closest to native collagen.