Localization of collagen prolyl hydroxylase to the hepatocyte : Studies in primary monolayer cultures of parenchymal cells from adult rat liver

Significant levels of prolyl hydroxylase activity (prolyl-glycyl-peptide, 2-oxoglutarate: oxygen oxidoreductase; EC 1.14.11.2) have been found in freshly isolated hepatocytes prepared from normal or regenerated adult rat liver and primary non-proliferating monolayer cultures of these cells. Four days after partial hepatectomy, the intact regenerated liver contained two times the normal level of prolyl hydroxylase activity. Freshly isolated hepatocytes contained 24% of the total prolyl hydroxylase activity in normal liver and 47% of that in regenerated liver. Upon incubation of hepatocytes for 24 h in a chemically defined culture medium containing insulin, prolyl hydroxylase activity rose 2- to 3-fold, and gradually declined during the next 48 h. The rise in prolyl hydroxylase activity was blocked by addition of cycloheximide to the culture medium. The presence of prolyl hydroxylase activity in hepatocyte cultures was not likely due to contamination with non-parenchymal liver cells. The latter cells contained less than 20% of the total enzyme activity recovered in all cells isolated from the liver. Furthermore, prolyl hydroxylase was localized by immunofluorescence uniformly to the hepatocytes in culture. Cultured hepatocytes converted [¹⁴C]proline to [¹⁴C]hydroxyproline at rates comparable to those reported for whole liver. However, only a small portion of the hydroxyproline containing product was present as collagen protein, suggesting its rapid degradation in culture. We conclude that the liver parenchymal cell may actively participate in collagen synthesis and possibly in collagen degradation.     

Regulation of collagen post-translational modification in transformed human and chick-embryo cells.

Changes in the regulation of collagen post-translational modification in transformed cells were studied in three established human sarcoma cell lines and in chick-embryo fibroblasts freshly transformed by Rous sarcoma virus. The collagens synthesized by all but one of these and by all the control human and chick-embryo cell lines were almost exclusively of types I and/or III. The relative rate of collagen synthesis and the amounts of prolyl hydroxylase activity and immunoreactive protein were markedly low in all the transformed human cell lines. The other enzymes studied, lysyl hydroxylase, hydroxylysyl galactosyltransferase and galactosylhydroxylysyl glucosyltransferase, never showed as large a decrease in activity as did prolyl hydroxylase, suggesting a more efficient regulation of the last enzyme than of the three others. The chick-embryo fibroblasts freshly transformed by Rous sarcoma virus differed from the human sarcoma cells in that prolyl hydroxylase activity was distinctly increased, whereas the decreases in immunoreactive prolyl hydroxylase protein and the three other enzyme activities were very similar to those in the simian-virus-40-transformed human fibroblasts. It seems possible that this increased prolyl hydroxylase activity is only a temporary phenomenon occurring shortly after the transformation, and may be followed by a decrease in activity later. The newly synthesized collagens of all the transformed cells that produced almost exclusively collagen types I and/or III had high extents of lysyl hydroxylation, and there was also an increase in the ratio of glycosylated to non-glycosylated hydroxylysine. The data suggest that one critical factor affecting modification is the rate of collagen synthesis, which affects the ratio of enzyme to substrate in the cell.     

The effect of hypoxia on collagen synthesis in cultured 3T6 fibroblasts and its relationship to the mode of action of ascorbate.

When exposed to low oxygen tension, in the absence of added ascorbic acid 3T6 mouse fibroblast cultures in late log phase respond by increased lactate production and increased hydroxylation of proline in nascent collagen, which is paralleled by an increase in prolyl hydroxylase activity. After 6 h recovery from the anoxic stimulus, however, cultures still yield more prolyl hydroxylase than controls, but the effect on hydroxylation of nascent collagen has disappeared. These observations help to dissect the dual role of ascorbate which can stimulate hydroxylation both by increasing the amount of active enzyme and by a cofactor-like role; in addition, these observations may be relevant to wound healing.     

Liver collagen hydroxylation in murine schistosomiasis.

The activity of procollagen prolyl hydroxylase was measured in fibrotic liver obtained from mice with hepatosplenic schistosomiasis, an animal model of the most prevalent form of human liver fibrosis. Measurable activity of prolyl hydroxylase in fibrotic liver supernatants was 47-fold higher than that of normal liver. The effect of prolyl hydroxylase inhibition on collagen synthesis in fibrotic liver slices was studied, using 8,9-dihydroxy-7-methyl benzo[b]quinolizinium bromide (GPA 1734). This compound was shown in other systems to inhibit prolyl and lysyl hydroxylations by iron chelation at concentrations which did not affect total protein synthesis. The formation of nondialyzable labelled hydroxyproline was inhibited by GPA 1734, 40, 70 and 95% at 30, 50 and 100 micrometer, respectively. Incorporation of proline into total liver protein was unaffected at 30 and 50 micrometer, but was inhibited 20% at 100 micrometer GPA 1734. Underhydroxylated collagen synthesized by liver slices with GPA 1734 was extracted with neutral salt solution and was subsequently hydroxylated with partially-purified prolyl hydroxylase to the same extent as control material synthesized in the absence of GPA 1734.     

Effect of L-3,4-dehydroproline on collagen synthesis and prolyl hydroxylase activity in mammalian cell cultures.

The incorporation of DL-3,4-dehydro[14C]proline into collagen and total protein of 3T3 cells occurred at approximately one-fifth the rate observed for L-[14C]proline. Addition of L-3,4-dehydroproline to the culture medium inhibited markedly the incorporation of [14C]glycine and L-[3H]lysine into the collagen of 3T3 cells, but there was only slight inhibition of the incorporation of the radiolabeled amino acids into total cellular proteins, indicating that the action of L-3,4-dehydroproline is specific for collagen. When 1 mM L-3,4-dehydroproline was added to the culture medium the [14C]hydroxyproline content was reduced 40% in the cell layer and 70% in the medium. The D isomer of 3,4-dehydroproline did not inhibit [14C]hydroxyproline formation. These findings indicate that L-3,4-dehydroline reduced the hydroxylation of the susceptible prolyl residues in the collagen molecule and the secretion of collagen from the cell. The reduction in the hydroxyproline content is probably related in part to a reduction in the activity of prolyl hydroxylase; when various mammalian cell cultures were exposed to 0.2 mM L-3,4-dehydroproline, the specific activity of prolyl hydroxylase was reduced markedly, while that of lysyl hydroproline, the specific activity of prolyl hydroxylase was reduced markedly, while that of lysyl hydroxylase was not affected. Under these conditions, cell growth and lactic dehydrogenase required protein synthesis. Removal of L-3,4-dehydroproline from the growth medium resulted in a time-dependent increase in the specific activity of prolyl hydroxylase.     

Liver collagen hydroxylation in murine schistosomiasis

The activity of procollagen prolyl hydroxylase was measured in fibrotic liver obtained from mice with hepatosplenic schistosomiasis, an animal model of the most prevalent form of human liver fibrosis. Measurable activity of prolyl hydroxylase in fibrotic liver supernatants was 47-fold higher than that of normal liver.The effect of prolyl hydroxylase inhibition on collagen synthesis in fibrotic liver slices was studied, using 8,9-dihydroxy-7-methyl benzo[b]quinolizinium bromide (GPA 1734). This compound was shown in other systems to inhibit prolyl and lysyl hydroxylations by iron chelation at concentrations which did not affect total protein synthesis. The formation of nondialyzable labelled hydroxyproline was inhibited by GPA 1734, 40, 70 and 95% at 30, 50 and 100 μM, respectively. Incorporation of proline into total liver protein was unaffected at 30 and 50 μM, but was inhibited 20% at 100μM GPA 1734. Underhydroxylated collagen synthesized by liver slices with GPA 1734 was extracted with neutral salt solution and was subsequently hydroxylated with partially-purified prolyl hydroxylase to the same extent as control material synthesized in the absence of GPA 1734.     

Proteolytic activity in human skin fibroblasts harvested with trypsin and its effect on measurements of collagen hydroxylase activities.

Lysates of human skin fibroblasts harvested without the use of trypsin do not contain detectable proteolytic activity, but when trypsin is used, lysates may contain activity equal to 10 ng of trypsin/10⁷ cells. The amount of cell lysate ordinarily examined for collagen prolyl and lysyl hydroxylase activity is sufficiently small that such amounts of trypsin have no observable effect on the unhydroxylated collagen substrate. Larger amounts of trypsin cause proteolysis of the unhydroxylated collagen substrate and a reduction of both prolyl and lysyl hydroxylation with lysyl hydroxylation more affected at low trypsin concentration than prolyl hydroxylation.     

Submicrosomal localization of prolyl hydroxylase from chick embryo limb bone.

A fraction greatly enriched in microsomes was prepared from chick embryo limb bone tissue homogenates by differential centrifugation in a high density solution of Metrizamide. This fraction was used to determine the submicrosomal localization of prolyl hydroxylase. At a low concentration (0.05%) of the non-ionic detergents Triton X-100 and Brij-35, 90 to 93% of prolyl hydroxylase activity was released from microsomes. Concentrations of Triton X-100 greater than 0.1% were required to solubilize the intrinsic membrane enzyme NADH-ferricyanide reductase and to release membrane-bound ribosomes, while Brij-35 did not extensively solubilize membrane components even at concentrations up to 0.4%. In addition, prolyl hydroxylase activity which could subsequently be released from microsomes by Brij-35 was relatively resistant to trypsin proteolysis at concentrations which removed more than 50% of the ribosomes and approximately 40% of the protein from microsomes. These results suggest that 90 to 93% of prolyl hydroxylase activity in connective tissue is located within the cisternae of the endoplasmic reticulum. Gel filtration of prolyl hydroxylase released from microsomes or found in the soluble fraction of limb bone homogenates revealed two peaks of activity corresponding to molecular weights of 230,000 and 450,000 to 500,000. The latter is twice the value reported for purified chick embryo prolyl hydroxylase. A fraction of the total prolyl hydroxylase activity (generally 20 to 35%) in microsome preparations could be measured in the absence of detergent, although the microsomal membrane should be impermeable to the large unhydroxylated collagen chains used as substrate. On the basis of experimental data, it was concluded that detergent-independent activity was most likely due to damaged microsomal membranes and that this damage was sufficient to allow substrate and trypsin to enter the cisternae but not to allow prolyl hydroxylase to be released.     

Collagen lysyl hydroxylation occurs within the cisternae of the rough endoplasmic reticulum.

Resolution of the heavy microsomal fraction of lung tissue by Ficoll density gradient centrifugation yielded a rough endoplasmic reticulum microsomal fraction containing the highest specific activity of detergent-released lysyl hydroxylase. This same microsomal fraction was previously shown to contain the highest specific activity of detergent-released prolyl hydroxylase activity. When hydroxylation was inhibited during the biosynthesis of collagen, this microsomal fraction contained lysine-rich, hydroxylysine-deficient, collagenase-digestible substrate that could be hydroxylated in the absence of detergent. The results indicate coordinate localization of both prolyl and lysyl hydroxylation reactions within the cisternae of the rough endoplasmic reticulum.     

Collagen biosynthesis. Characterization of subcellular fractions from embryonic chick fibroblasts and the intracellular localization of protocollagen prolyl and protocollagen lysyl hydroxylases

1. Subcellular fractions of freshly isolated matrix-free embryonic chick tendon and sternal cartilage cells have been characterized by chemical analysis, electron microscopy and the location of specific marker enzymes. These data indicate the fractions to be of a high degree of purity comparable with those obtained from other tissues, e.g. liver and kidney. 2. When homogenates were assayed for protocollagen prolyl hydroxylase and protocollagen lysyl hydroxylase activities, addition of Triton X-100 (0.1%, w/v) was found to stimulate enzyme activities by up to 60% suggesting that the enzymes were probably membrane-bound. 3. Assay of subcellular fractions obtained by differential centrifugation for protocollagen prolyl hydroxylase activity indicated the specific activity to be highest in the microsomal fraction. Similar results were obtained for protocollagen lysyl hydroxylase activity. 4. Submicrosomal fractions obtained by discontinuous sucrose-gradient centrifugation were assayed for the two enzymes and protocollagen prolyl hydroxylase and protocollagen lysyl hydroxylase were found to be associated almost exclusively with the rough endoplasmic reticulum fraction in both tendon and cartilage cells.     

Characterization of collagen hydroxylysyl glycosyltransferases as mainly intramembranous microsomal enzymes

The localization of collagen hydroxylysine galactosyl- and galactosyl-hydroxylysine glucosyltransferases in purified chick embryo bone microsomes was studied by differential solubilization with nonionic detergents. Brij-35 (polyoxyethylene 25-lauryl ether) which selectively releases intracisternal proteins, and Triton X-100, whose specificity varies with its concentration, were used in the presence or absence of high ionic strength NaCl. These methods were used previously to characterize prolyl hydroxylase as intracisternal and lysyl hydroxylase as mainly intramembranous. The distribution of both glycosyltransferases within microsomes was similar to that of lysyl hydroxylase; approximately 70-80% of their activities are intramembranous with the remainder intracisternal. Collagen hydroxylysine glucosyltransferase differed from prolyl and lysyl hydroxylase and the galactosyltransferase in that its activity in vitro was apparently inhibited by membrane vesicles, even in the presence of detergents at concentrations which permeabilize the membrane. Accurate measurement of its activity could be achieved only by its separation from vesicles after detergent treatment. The common location of the major portion of lysyl hydroxylase and the glycosyltransferase activities suggests that they may act as a multienzyme complex to preferentially modify certain lysyl residues in nascent procollagen chains as they traverse the membrane of the endoplasmic reticulum. Since these enzymes do not act on helical collagen, their physical separation from prolyl hydroxylase may ensure that modifications of lysine residues occur prior to formation of hydroxyproline, which stabilizes the helical form.     

Collagen accumulation can continue with decreased prolyl hydroxylase activity in the liver

A single dose of dimethylnitrosamine dose-relatedly increased total hepatic hydroxyproline content in rats 14 days after the dosing. In cases of 35 mg/kg of dimethylnitrosamine, it increased rapidly to 1.7 times the normal level within 14 days. This increase persisted thereafter until 84 days. Hepatic collagen prolyl hydroxylase activity was 1.8 times the normal level by the fourth day after the dosing but normalized within 14 days. It decreased further to levels significantly lower than those in normal rats at 28, 56 and 84 days. Serum glutamic pyruvic transaminase activity was about 13 times the normal level at 2 days and normalized after 7 days. On histology, fiber developed in necrotic areas around the central veins after 7 days and remained after 28 days when the necrosis had already disappeared. These results suggest that abnormal collagen can continue to increase in the state of decreased collagen prolyl hydroxylase activity in the liver.     

Immunological characterization of lysyl hydroxylase, an enzyme of collagen synthesis

Antibodies to pure lysyl hydroxylase from whole chick embryos were prepared in rabbits and used for immunological characterization of this enzyme of collagen biosynthesis. In double immunodiffusion a single precipitation line was seen between the antiserum and crude or pure chick-embryo lysyl hydroxylase. The antiserum effectively inhibited chick-embryo lysyl hydroxylase activity, whether measured with the biologically prepared protocollagen substrate or a synthetic peptide consisting of only 12 amino acids. This suggests that the antigenic determinant was located near the active site of the enzyme molecule. Essentially identical amounts of the antiserum were required for 40% inhibition of the same amount of lysyl hydroxylase activity units from different chick-embryo tissues synthesizing various genetically distinct collagen types. In double immunodiffusion a single precipitation line of complete identity was found between the antiserum and the purified enzyme from whole chick embryos and the crude enzymes from chick-embryo tendon, cartilage and kidneys. These results do not support the hypothesis that lysyl hydroxylase has collagen-type-specific or tissue-specific isoenzymes with markedly different specific activities or immunological properties. The antibodies to chick-embryo lysyl hydroxylase showed a considerable degree of species specificity when examined either by activity-inhibition assay or by double immuno-diffusion. Nevertheless, a distinct, although weak, cross-reactivity was found between the chick-embryo enzyme and those from all mammalian tissues tested. The antiserum showed no cross-reactivity against prolyl 3-hydroxylase, hydroxylysyl galactosyl-transferase or galactosylhydroxylysyl glucosyltransferase in activity-inhibition assays, whereas a distinct cross-reactivity was found against prolyl 4-hydroxylase. Furthermore, antiserum to pure prolyl 4-hydroxylase inhibited lysyl hydroxylase activity. These findings suggest that there are structural similarities between these two enzymes, possibly close to or at their active sites.     

Phosphorylation of pig brain microtubule proteins. General properties and partial characterization of endogenous substrate and cyclic AMP-dependent protein kinase.

Collagen synthesis and the activities of prolyl hydroxylase, lysyl hydroxylase, collagen galactosyltransferase and collagen glucosyltransferase were studied in isolated chick-embryo tendon cells after the administration of cortisol acetate to the chick embryos. When the steroid was injected 1 day before isolation of the tendon cells, collagen synthesis was decreased, even though the enzyme activities were not changed. When cortisol acetate was given as repeated injections over a period of 4 days, both collagen synthesis and the enzyme activities decreased. The hydroxylase activities decreased even more than the two collagen glycosyltransferase activities, both in isolated cells and in whole chick embryos. The amount of prolyl hydroxylase protein diminished to the same extent as the enzyme activity, indicating that cortisol acetate inhibits enzyme synthesis. The inhibitory effect of cortisol acetate on collagen synthesis and on the enzyme activities was partially reversible in 3 days. Total protein synthesis was completely restored within this time. Only massive doses of cortisol acetate inhibited collagen synthesis in vitro. Additional experiments indicated that cortisol acetate did not decrease the rate of the enzyme reactions when added directly to the enzyme incubation mixtures. The results suggest that cortisol acetate decreases collagen synthesis both by its direct effect on collagen polypeptide-chain synthesis and by decreasing the activities of enzymes involved in post-translational modifications.     

Recombinant expression of hydroxylated human collagen in Escherichia coli

Collagen is the most abundant protein in the human body and thereby a structural protein of considerable biotechnological interest. The complex maturation process of collagen, including essential post-translational modifications such as prolyl and lysyl hydroxylation, has precluded large-scale production of recombinant collagen featuring the biophysical properties of endogenous collagen. The characterization of new prolyl and lysyl hydroxylase genes encoded by the giant virus mimivirus reveals a method for production of hydroxylated collagen. The coexpression of a human collagen type III construct together with mimivirus prolyl and lysyl hydroxylases in Escherichia coli yielded up to 90 mg of hydroxylated collagen per liter culture. The respective levels of prolyl and lysyl hydroxylation reaching 25 % and 26 % were similar to the hydroxylation levels of native human collagen type III. The distribution of hydroxyproline and hydroxylysine along recombinant collagen was also similar to that of native collagen as determined by mass spectrometric analysis of tryptic peptides. The triple helix signature of recombinant hydroxylated collagen was confirmed by circular dichroism, which also showed that hydroxylation increased the thermal stability of the recombinant collagen construct. Recombinant hydroxylated collagen produced in E. coli supported the growth of human umbilical endothelial cells, underlining the biocompatibility of the recombinant protein as extracellular matrix. The high yield of recombinant protein expression and the extensive level of prolyl and lysyl hydroxylation achieved indicate that recombinant hydroxylated collagen can be produced at large scale for biomaterials engineering in the context of biomedical applications.     

Inhibitory effects of glucocorticoids on collagen synthesis by mouse sponge granulomas and granuloma fibroblasts in culture.

The basis for the glucocorticoid-mediated decrease in tissue collagen was studied in mouse granulomas and in primary granuloma fibroblast cultures. Injection of mice for 12 days with dexamethasone (0.35 mg/kg body weight) resulted in a 50--70% inhibition of collagen synthesis and accumulation in polyvinyl sponge-induced granulomas whereas total protein synthesis was inhibited by only about 25%. The decreased collagen content of the granuloma was accounted for by both a reduced fibroblast number and diminished synthesis per cell. Growth rates, total protein synthesis and collagen synthesis were the same in granuloma fibroblast cultures derived from control or steroid-treated mice. However, addition of 3.10(-7) M hydrocortisone to the culture medium caused a 30--50% inhibition of both collagen and non-collagen protein synthesis in firbroblasts from either source. These inhibitory effects were dose- and time-dependent with a lag time of 12--24 h. Prolyl hydroxylase activity was reduced both in sponge granulomas from glucocorticoid-treated mice and in hydrocortisone-treated fibroblast cultures. However, protein synthesis was inhibited to the same extent as the inhibition of prolyl hydroxylase activity and there was no effect on peptidyl prolyl hydroxylation. These results indicate that the glucocorticoid-induced reduction of collagen synthesis and accumulation observed in mouse granulomas and primary granuloma fibroblast cultures is not specific for this protein. Furthermore, glucocorticoid-induced inhibition of collagen synthesis cannot be attributed to underhydroxylation of collagen prolyl residues.     

Prolyl hydroxylase activity in L-929 fibroblasts incubated with and without ascorbate

An improved procedure was used to assay prolyl hydroxylase activity in both early-log and late-log L-929 fibroblasts grown on plastic surfaces. When 40 μg/ml of ascorbate was added to early-log phase cultures, the rate of hydroxy-[¹⁴C] proline synthesis increased 2-fold within 4 h, but there was no change in prolyl hydroxylase activity per cell. The results indicated therefore that ascorbate did not “activate” prolyl hydroxylase in the sense of converting inactive enzyme protein to active enzyme protein. Instead ascorbate appeared to increase hydroxyproline synthesis in early-log L-929 fibroblasts because the prolyl hydroxylase reaction in such cells was limited by the availability of ascorbate or a similar cofactor. When 40 μg/ml of ascorbate was added to late-log phase cultures, there was essentially no effect on the rate of hydroxyl[¹⁴C]-proline synthesis or prolyl hydroxylase activity. The late-log phase cells, however, contained three times more enzyme activity and about two times more immuno-reactive enzyme protein than early-log phase cells. In addition, the rate of protein synthesis per cell in late-log phase cells was only one-tenth the rate in early-log phase cells. The results suggested that as the cells grew to confluency, collagen polypeptides were more completely hydroxylated in part because the rate of polypeptide synthesis decreased and at the same time prolyl hydroxylase activity per cell increased. The results appear to provide an alternate explanation for previous observations on the effects of ascorbate and “crowding” on hydroxy[roline synthesis in cultures of L-929 fibroblasts.     

Collagen synthesis by cultured skin fibroblasts from siblings with hydroxylysine-deficient collagen.

It has been previously shown that dermis from subjects with hydroxylysine-deficient collagen contains approximately 5% of normal levels of hydroxylysine and sonicates of skin fibroblasts contain less than 15% of normal levels of collagen lysyl hydroxylase activity. However, cultures of dermal fibroblasts from two siblings with hydroxylysine-deficient collagen (Ehlers-Danlos Syndrome Type VI) compared to fibroblasts from normal subjects synthesize collagen containing approximately 50% of normal amounts of hydroxylysine. The lysyl hydroxylase deficient cultures synthesize both Type I and Type III collagen in the same proportion as control cultures. Both alpha 1(I) and alpha 2 chains are similarly reduced in hydroxylysine content. Collagen prolyl hydroxylation by normal collagen lysyl hydroxylation is the same with or without ascorbate supplementation. In mutant cells the rate of prolyl hydroxylation measured after release of inhibition by alpha, alpha'-dipyridyl is the same as in control cells. The rate of lysyl hydroxylation is reduced in mutant cells but only to approximately 50% of normal.     

Lack of collagen type specificity for lysyl hydroxylase isoforms

Lysyl hydroxylase is the enzyme catalyzing the formation of hydroxylysyl residues in collagens. Large differences in the extent of hydroxylysyl residues are found among collagen types. Three lysyl hydroxylase isoenzymes (LH1, LH2, LH3) have recently been characterized from human and mouse tissues. Nothing is known about the distribution of these isoforms within cells or whether they exhibit collagen type specificity. We measured mRNA levels of the three isoforms, as well as the mRNAs of the main collagen types I, III, IV, and V and the alpha subunit of prolyl 4-hydroxylase, another enzyme involved in collagen biosynthesis, in different human cell lines. Large variations were found in mRNA expression of LH1 and LH2 but not LH3. Immunoblotting was utilized to confirm the results of Northern hybridization. The levels of mRNA of LH1, LH2, and the alpha subunit of prolyl 4-hydroxylase showed significant correlations with each other. The LH3 mRNA levels did not correlate with those of LH1, LH2, or the alpa subunit of prolyl 4-hydroxylase, clearly indicating a difference in the regulation of LH3. No correlation was observed between LH isoforms and individual collagen types, indicating a lack of collagen type specificity for lysyl hydroxylase isoforms. Our observations suggest that LH1, LH2, and the alpha subunit of prolyl 4-hydroxylase are coregulated together with total collagen synthesis but not with the specific collagen types and indicate that LH3 behaves differently from LH1 and LH2, implying a difference in their substrates. These observations set the basis for further studies to define the functions of lysyl hydroxylase isoforms.