P3h3-null and Sc65-null Mice Phenocopy the Collagen Lysine Under-hydroxylation and Cross-linking Abnormality of Ehlers-Danlos Syndrome Type VIA*

Tandem mass spectrometry was applied to tissues from targeted mutant mouse models to explore the collagen substrate specificities of individual members of the prolyl 3-hydroxylase (P3H) gene family. Previous studies revealed that P3h1 preferentially 3-hydroxylates proline at a single site in collagen type I chains, whereas P3h2 is responsible for 3-hydroxylating multiple proline sites in collagen types I, II, IV, and V. In screening for collagen substrate sites for the remaining members of the vertebrate P3H family, P3h3 and Sc65 knock-out mice revealed a common lysine under-hydroxylation effect at helical domain cross-linking sites in skin, bone, tendon, aorta, and cornea. No effect on prolyl 3-hydroxylation was evident on screening the spectrum of known 3-hydroxyproline sites from all major tissue collagen types. However, collagen type I extracted from both Sc65−/− and P3h3−/− skin revealed the same abnormal chain pattern on SDS-PAGE with an overabundance of a γ112 cross-linked trimer. The latter proved to be from native molecules that had intramolecular aldol cross-links at each end. The lysine under-hydroxylation was shown to alter the divalent aldimine cross-link chemistry of mutant skin collagen. Furthermore, the ratio of mature HP/LP cross-links in bone of both P3h3−/− and Sc65−/− mice was reversed compared with wild type, consistent with the level of lysine under-hydroxylation seen in individual chains at cross-linking sites. The effect on cross-linking lysines was quantitatively very similar to that previously observed in EDS VIA human and Plod1−/− mouse tissues, suggesting that P3H3 and/or SC65 mutations may cause as yet undefined EDS variants.

The collagen family of proteins has evolved into many different genes and gene translational products each with variably regulated post-translational modifications (1,2). Even within a single genetic type of collagen, the post-translational quality of the protein is known to be regulated with a high degree of tissue specificity (3,4). The functional significance of this post-translational variability is still not fully understood, although for fibril-forming collagens the number, placement, and chemistry of covalent intermolecular cross-links seem to be critically important regulators of tissue material properties and function (5)(6)(7).
In the last decade, new insights on the significance of a relatively rare collagen modification, prolyl 3-hydroxylation, came from the discovery that recessive forms of osteogenesis imperfecta (OI) 2 are caused by biallelic mutations in prolyl 3-hydroxylase 1 (P3H1; Lepre1), CRTAP (Leprel3), or CypB (PPIB) (8). These proteins, which form a P3H1 enzyme complex, act on nascent collagen chains in the ER (9). Mutations in at least 6 further genes that encode either enzymes (e.g. PLOD2 encoding lysyl hydroxylase-2 (LH2) (10)) or chaperones (e.g. FKBP10 encoding FKBP65 (11,12)), needed for collagen modification, folding, transport, and normal mineralization, have been shown to cause OI variants. We have gained insights on the disease-causing mechanisms by analyzing tissue collagens from OI patients and mouse models of OI. For example, P3H1, CRTAP, and PPIB mutations all cause telopeptide lysine overhydroxylation (5), whereas PLOD2 or FKBP10 mutations cause telopeptide lysine under-hydroxylation (10 -12). A common effect from all these mutations is an altered collagen cross-linking chemistry. These findings suggest an interplay in the ER between the prolyl 3-hydroxylation complex and the lysyl hydroxylation machinery.
Human mutations in P3H2 (Leprel1) have been shown to cause the eye disorder high myopia (13)(14)(15). Two different P3H2 knock-out mice have so far been generated by separate laboratories, the first was embryonic lethal (16), whereas the second had no obvious phenotype (17). Mass spectral analysis of multiple tissues from the latter P3H2 knock-out mouse revealed that collagen types I, II, V, and IV lack 3-hydroxypro-line at all sites other than the few sites acted on by P3H1 (17). But so far, neither substrate specificity nor disease-associated mutations have been identified for P3H3. Recombinant P3H3 protein expressed in bacteria was reported to form insoluble aggregates with no detectable hydroxylating activity (18). The fifth remaining member of the gene family, synaptonemal complex 65 (Sc65), is a close homologue of Crtap with an apparent function in bone homeostasis (19), as well as a suspected role as an autoimmune target in membranous nephropathy (20). We recently reported suppressed collagen lysine hydroxylation in the Sc65-null mouse and evidence for complex formation between Sc65, P3H3, and potentially lysyl hydroxylase-1 (LH1) and CypB (21).
To understand more about the collagen substrate specificity of P3H3 and the function of Sc65, we screened tissues from Sc65-and P3h3-null mice using tandem mass spectrometry for collagen post-translational effects. The aim was to see if any known sites of 3-hydroxyproline in the various collagen types were under-hydroxylated and also if other post-translational modifications, in particular lysine modification and collagen cross-linking chemistry were affected. The results showed that both P3h3-and Sc65-null mice displayed a similar pattern of severe under-hydroxylation of lysines, particularly at crosslinking sites in type I collagen from bone, skin, dentin, cornea, and aorta. No loss of prolyl 3-hydroxylation was evident at any site accessible to tryptic peptide mass spectral screening of collagen types I, II, IV, and V from several tissue sources. The findings point to a complex evolution and interplay between the proteins forming the machinery of collagen prolyl 3-hydroxylation and lysyl hydroxylation in the ER.

Results
Generation of Sc65 and P3H3 Knock-out Mice-The murine Leprel4 (Sc65) gene was inactivated by homologous recombination in ES cells using two approaches. The first mouse model was established via gene trap insertional mutagenesis (19), whereas the second was generated by introducing flanking loxP sites in the Leprel4 gene (floxed) and then bred with a mouse ubiquitously expressing Cre-recombinase (21). Generation and initial characterization of each of these Sc65-null mouse models has been described (19,21). The murine Leprel2 (P3h3) gene was inactivated by homologous recombination in ES cells using BAC technology (Fig. 1A). Cross-breeding of heterozygous mice (P3h3 ϩ/Ϫ ) produced the null genotypes. Immunoblot analysis using a polyclonal P3H3 antibody against fractionated renal protein extracts showed absence of protein expression in the P3h3 Ϫ/Ϫ mice (Fig. 1B). A 4 M guanidine HCl extract of fetal human cartilage was used as the positive control.
Both homozygous mouse strains, Sc65 Ϫ/Ϫ and P3h3 Ϫ/Ϫ , were viable and showed no skeletal growth defects grossly or histologically or any other abnormalities when observed in the cage environment. P3h3 Ϫ/Ϫ skin sections on H&E staining (Fig.  1C) showed some evidence of collagen fabric fragility (Fig. 1C), similar to that observed in Sc65 Ϫ/Ϫ skin sections (21). Furthermore, on dissection, the skin of both homozygous null strains appeared to lack the structural integrity of normal mouse skin. Skin from Sc65 Ϫ/Ϫ mice also proved on mechanical testing to fail at lower loads (21). Unless otherwise stated, the WT mouse genotype used in this study was Sc65 ϩ/ϩ . Tissue collagens from the heterozygote P3h3 ϩ/Ϫ mouse, which were determined to be indistinguishable from Sc65 ϩ/ϩ mouse (WT) by collagen SDS-PAGE and mass spectral analysis, were also used as a control in this study where indicated.
Collagen Chain Profile on SDS-PAGE-Acid-extracted collagens from skin, bone, and tendon were resolved on SDS-PAGE to assess potential tissue-specific differences in the collagen cross-linking profiles. Reproducible and similar differences in collagen chain patterns were evident from skin of both Sc65 Ϫ/Ϫ and P3h3 Ϫ/Ϫ mice compared with their respective WT littermates (the latter patterns being indistinguishable). Significantly, a band we previously identified as the chain trimer, ␥ 112 (12,22), was reproducibly more prominent in the null mouse collagen extracts relative to the ␣ and ␤ bands (Fig. 2). The ␤ 12 component from both null strains was also noticeably depleted compared with WT in proportion with the increase in ␥ 112 . From normal skin and tendon, the ␤-dimers are known to be the result of intramolecular aldol cross-link formation between allysines in ␣1-␣1 and ␣1-␣2 N-telopeptides. The shift from ␤ 12 to ␥ 112 observed with the null mice suggested an effect on intermolecular cross-linking. We reported a similar banding pattern from fibromodulin null mouse tendons and concluded FIGURE 1. Generation of P3h3 ؊/؊ mice. A, the Leprel2 gene was inactivated by homologous recombination in ES cells using BAC technology. Mice were maintained on a C57/BL6J background. B, immunoblot analysis using a polyclonal P3H3 antibody against total protein extracts from whole kidneys of WT and P3h3 Ϫ/Ϫ mice supported the loss of protein in the knock-out mice. Human fetal cartilage extract was loaded as a positive control marker for P3H3. C, representative H&E images of skin sections from P3h3 Ϫ/Ϫ and WT mice.
the cause was ␣1-␣1 C-telopeptide aldol formation due to enhanced lysyl oxidase oxidation of ␣1-C-telopeptide lysines (22). Further studies suggested that fibromodulin interacts with lysyl oxidase supporting this conclusion (23). Interestingly, the shift in band intensity in the current mice was most prominent in skin, less so in tendon, and absent in bone collagen extracts.
Lack of Effect on Prolyl 3-Hydroxylation-Tandem mass spectrometry was used as a screen to survey the degree of prolyl 3-hydroxylation at all known sites we could readily access as tryptic peptides from collagen types I, II, IV, and V. No significant evidence of suppression was observed at any site in Sc65 Ϫ/Ϫ or P3h3 Ϫ/Ϫ tissues (Table 1 details the findings). This was surprising, particularly for P3h3 Ϫ/Ϫ , which is a close homolog of P3h1 and P3h2, including a potentially active catalytic domain. Both P3h1 and P3h2 have evolved preferred GPP substrate sequences in a range of collagen types (17,24,25).
Identification of the Molecular Structure of the ␥ 112 Trimer-The molecular nature of the ␥ 112 trimer was investigated using in-gel trypsin mass spectrometry. Mass spectrometric analysis revealed a prominent dimeric ␣1(I) C-telopeptide with its cross-linking residue fitting exactly an ␣,␤-unsaturated allysine aldol (Fig. 3A). As previously shown, trypsin does not cleave after lysines that have been converted to lysine aldehydes, resulting in 6-residue longer tryptic telopeptides cleaved after arginine (22). Notably, MS after in-gel trypsin digest of the ␥ 112 trimer revealed no monomeric ␣1(I) C-telopeptide, only the cross-linked dimer. MS-MS confirmed the identity of the dimeric cross-linked tryptic peptide (Fig. 3B). The ϩ16 Da mass ladder given by the parent ion is the result of variable 4-hydroxylation of the Yaa position Pro in the C-terminal GPP of the triple-helical domain of each ␣1-chain of the dimer.
Tandem Mass Spectrometry Targeted at ␣1(I) and ␣2(I) Cross-linking Lysines Shows Under-hydroxylation-Lysine hydroxylation at the cross-linking site, residue Lys-87, was investigated using tandem mass spectrometry of type I collagen ␣1(I) chains extracted from bone, skin, tendon, aorta, and cornea. A similar pattern of reduced hydroxylation was observed from both Sc65 Ϫ/Ϫ and P3h3 Ϫ/Ϫ mouse tissues, but the effect was consistently greater with Sc65 Ϫ/Ϫ mice. As previously reported for the Sc65 Ϫ/Ϫ mouse, the post-translational shift was more pronounced in skin at helical domain cross-linking residue Lys-87 ( Fig. 4) compared with bone. Type I collagen from all soft tissues (skin, tendon, and aorta) showed almost complete loss of post-translational modification of Lys-87 ( Table 2). As in the Sc65 Ϫ/Ϫ mouse, all helical cross-linking sites appear to be under-hydroxylated in the P3h3 Ϫ/Ϫ mouse. Indeed, the ␣1(I) Lys-930 residue, which is completely hydroxylated in P3h3 ϩ/Ϫ skin, is unmodified in the P3h3 Ϫ/Ϫ mouse skin (Fig. 5). A similar loss of hydroxylation was found at Lys-930 in skin, bone, and tendon from both P3h3 Ϫ/Ϫ and Sc65 Ϫ/Ϫ mice. WT tendon was unique among soft tissues with 100% hydroxylated and FIGURE 2. SDS-PAGE reveals altered cross-linking in skin type I collagen from P3h3 ؊/؊ and Sc65 ؊/؊ mice. As visualized by 6% SDS-PAGE, type I collagen from P3h3 Ϫ/Ϫ and Sc65 Ϫ/Ϫ mouse skin acid extracts reproducibly yielded more ␥ 112 relative to WT. Accompanying this increased band intensity in knock-out mice skin is the decrease in ␤ 12 relative to WT mice. This shift in banding pattern is not observed in acid extracts of bone and tendon.

TABLE 1 Comparison of 3Hyp occupancy in collagens from Sc65 -/-, P3h3 -/-, and control mouse tissues
Percentage of 3Hyp at each major substrate site in type I collagen in tendon (T), bone (B), skin (Sk), kidney (K), cornea (Co), and sclera (Sc); type II collagen in cartilage (Cart), type IV collagen in kidney (K), and type V collagen in bone (B). Each percentage was determined based on the ratio of the m/z peaks of each post-translational variant from a minimum of three mice unless indicated. ( a two mice; b one mouse).
non-glycosylated Lys-87 on both ␣-chains of type I collagen. The effect on Lys-87 hydroxylation was greater in Sc65 Ϫ/Ϫ tendon compared with P3h3 Ϫ/Ϫ tendon. The data are compiled in Table 2.
Type V Collagen Under-hydroxylation at Cross-linking Sites Is Restricted to Soft Tissue-The skin dominated reduction in hydroxylation and glycosylation of the helical domain crosslinking lysine 87 was also evident in type V collagen but to a lesser extent than in type I (Table 3). Lys-87 of ␣1(V) and ␣2(V) from WT skin was fully hydroxylated and glycosylated as glcgal-Hyl. A mild decrease in these modifications is seen from skin of Sc65 Ϫ/Ϫ and P3h3 Ϫ/Ϫ mice. As with type I collagen, the effect appeared to be more pronounced in the Sc65 Ϫ/Ϫ mouse. In contrast, no post-translational variances were observed in the type V collagen ␣-chains from bone between WT and Sc65 Ϫ/Ϫ or P3h3 Ϫ/Ϫ (Table 3).
Altered Aldimine Divalent Cross-linking in Mutant Skin-Bacterial collagenase digests of borohydride-reduced skin were analyzed by mass spectrometry to target the pool of aldimine divalent cross-links in the whole tissue collagen. This approach is useful for tracking structural intermediates in the lysine aldehyde cross-linking pathway. The LC-MS profile of the C-telopeptide to Lys-87 cross-linked peptide from collagenase-digested WT skin shows a fully glycosylated reduced aldimine cross-linked structure between ␣1(I) Lys-87 and an ␣1(I) C-telopeptide lysine (Fig. 6A). This is consistent with the predominance of glcgal-Hyl at Lys-87 in the linear tryptic peptide from the ␣1(I) chain extracted from WT skin. From P3h3 Ϫ/Ϫ mouse skin, about equal amounts of the fully glycosylated and a nonglycosylated reduced aldimine cross-link structure were recovered (Fig. 6, B and C). This finding suggests that glcgal-Hyl at Lys-87 is favored chemically and sterically over Lys and nonglycosylated Hyl as an aldimine end product on the lysine aldehyde cross-linking pathway, as only ϳ20% of the ␣1(I) chains from P3h3 Ϫ/Ϫ skin were glycosylated at Lys-87 ( Fig. 4; Table 2). The total ion current, an index of the amount of the crosslinked peptide in the MS scan, was ϳ10 times higher for the WT aldimine product compared with the equivalent P3h3 Ϫ/Ϫ aldimine products. Although this should be considered only semi-  In-gel trypsin digests typically undergo complete oxidation of methionine residues yielding methionine sulfoxide. The trypsin-generated peptide is shown with P* indicating 4Hyp, M* indicating methionine sulfoxide, and -galglc indicating glucosyl-galactosyl-.
quantitative, it suggests that the mutant skin has significantly fewer aldimine cross-links at this site than WT skin. The crosslink between ␣1(I) Lys-930 and the ␣1(I)N-telopeptide was fully hydroxylated in WT skin (hydroxylysino-norleucine on reduction) (Fig. 7A); however, in P3h3 Ϫ/Ϫ mouse skin only the non-hydroxylated aldimine cross-link was identified (lysinonorleucine on reduction) (Fig. 7B).
Pyridinoline (HP and LP) Cross-link Analysis-As an independent downstream measure of the effect of the loss of Sc65 or P3H3 on lysine cross-linking, pyridinoline cross-links were quantified after acid hydrolysis of samples of bone, aorta, lung, cartilage, and dentin collagen from each genotype ( Table 4). The results showed a significantly lower ratio of HP/LP in bone, aorta, and lung in both knock-out mouse strains. A similar reversal in the HP/LP ratio of the cross-linking pyridinolines of bone and other tissues is seen in patients with Ehlers-Danlos syndrome (EDS) VI (26). A comparable but less marked shift in the HP/LP ratio was found in dentin and cartilage samples from the Sc65 Ϫ/Ϫ mice. No genotypic significance can be attached to variations in total content (HP ϩ LP mol/mol of collagen) between samples as shown in Table 4, which may be due to sampling site variability.

Discussion
Although both construct strains of Sc65 Ϫ/Ϫ (gene trap and floxed) and P3h3 Ϫ/Ϫ mice were indistinguishable by gross physical appearance from their WT littermates, the mutant skins lacked the structural integrity of normal skin on dissection. It is notable that increased skin fragility and low bone mass were quantified in the Sc65 Ϫ/Ϫ mouse strain compared with WT littermates (21). On mass spectrometric analysis, type I collagen from both Sc65 Ϫ/Ϫ and P3h3 Ϫ/Ϫ tissues revealed

Comparison of type I collagen lysine 87 post-translational modifications between tissues
Modifications on lysine 87 from both ␣-chains of type I collagen in WT, Sc65 Ϫ/Ϫ , and P3h3 Ϫ/Ϫ mice were measured using mass spectrometry in skin, bone, tendon, and aorta. Lysine modifications include unmodified Lys, hydroxylysine (Hyl), galactosyl-hydroxylysine (G-Hyl), and glucosylgalactosyl-hydroxylysine (GG-Hyl). Each percentage was determined based on the ratio of the m/z peaks of each post-translational variant from a minimum of three mice unless indicated. severe under-hydroxylation of lysine residues at triple-helical domain cross-linking sites for skin, tendon, cornea, and aorta. In addition to under-hydroxylation, the residual Hyl87 was under-glycosylated compared with WT for skin, aorta, and cornea. Type V collagen from null mouse skin and aorta exhibited a decrease in hydroxylation/glycosylation at Lys-87 compared with WT; however, no change was detected in type V collagen from null-bone versus WT bone. As collagen type V acts as a quantitatively minor but essential template for type I collagen heterofibril polymerization, it is possible that even a minor disturbance in collagen V cross-linking could be amplified in the mature fibril. The present phenocopied effects on collagen lysine hydroxylation can best be explained if Sc65 and P3H3 form a complex in the ER that is required to facilitate access of LH1 to substrate lysine residues on newly made collagen chains. We proposed an interaction complex that at least transiently involves LH1, CypB, SC65, and P3H3 (21). Similar to the P3H1/CRTAP/CypB enzyme complex, which is required to 3-hydroxylate Pro-986, the primary site of 3Hyp in type I collagen, all four ER components LH1, CypB, SC65, and P3H3 are essential for normal triple-helix lysine hydroxylation. The requirement of CypB for collagen post-translational modification became clear from the effects of PPIB mutations that cause severe osteogenesis imperfecta. They result in loss of prolyl-hydroxylation at Pro-986 (27) and, in CypB knock-out mice, defective lysine hydroxylation in bone (28) and tendon (29) collagens. Furthermore, we know that FKBP65, another peptidyl prolyl isomerase and ER chaperone, can regulate the activity of LH2 the predicted telopeptide lysine hydroxylase (12,30,31).

Modification
Tissues from both P3h2 Ϫ/Ϫ and Crtap Ϫ/Ϫ mice showed no effect on Lys-87 hydroxylation/glycosylation, ruling out any potential redundancy or compensatory function between P3H3, P3H2, and P3H1 subunits in the ER. 3 In contrast to the functionally interdependent complex P3H1/CRTAP required for Pro-986 3-hydroxylation, P3H3/Sc65 appears to be redundant as a prolyl 3-hydroxylase. We had anticipated that P3H3 would act as a prolyl 3-hydroxylase somewhere on collagen, but despite a broad screen of candidate sites in the common collagens no evidence of an affected site was found (Table 1). Given the recent identification of many low occupancy 3-hydroxyproline sites in type IV (32) and type V (33) collagens, and a long list of genetic types of collagen with GPP triplets, it is possible that a proline substrate for P3H3 remains to be identified. Equally possible is lost enzyme activity over the course of evolution but a retained collagen chaperone function in the ER that optimizes LH1-mediated lysine hydroxylation particularly at helical domain cross-linking sites.
The pathogenic significance of the loss of Pro-986 3-hydroxylation in recessive cases of osteogenesis imperfecta caused by P3H1, CRTAP, or CypB mutations is still not clear, although absence of the complex as a procollagen chaperone rather than absence of 3Hyp is indicated (34,35). In this context, the overhydroxylation of telopeptide lysines seen in the bone collagen of OI patients and mouse models caused by mutations in P3H1, CRTAP, and CypB should be noted.
Hydroxylysines are functionally important in collagen for covalent cross-linking. In type I collagen four triple-helical domain lysine sites (␣1(I) Lys-87 and Lys-930; ␣2(I) Lys-87 and Lys-933) can interact with C-and N-telopeptide lysines after the latter have been converted to aldehyde residues by lysyl oxidase. Tissue-specific differences in collagen cross-linking chemistry and maturation are largely determined in the ER by the lysine hydroxylation state of these specific lysines (Fig. 8A).
Based on the present findings, what mechanism can explain the altered pattern (Fig. 2) of cross-linked ␤and ␥-chains seen on SDS-PAGE of skin collagen from Sc65 Ϫ/Ϫ and P3h3 Ϫ/Ϫ mice? We believe the latter effect is caused by the lack of glcgal-Hyl at Lys-87 in both ␣-chains of skin type I collagen. In WT skin, aorta and cornea (Ͼ90%), type I collagen ␣1(I) chains all have fully glycosylated glcgal-Hyl at Lys-87 to which ␣1(I) C-telopeptide allysines in neighboring four-dimensional staggered molecules normally interact to form aldimine cross-links. When ␣1(I) and ␣2(I) Lys-87 are not hydroxylated and so cannot be glycosylated the interaction chemistry changes. If in the same molecule both ␣1-chain C-telopeptide lysines are converted to aldehydes, then they interact to form an intramolecular aldol cross-link in lieu of the typical intermolecular glyco-3 D. M. Hudson and D. R. Eyre, unpublished data.

TABLE 3 Summary of posttranslational variances in linear cross-linking lysine 87 from type V collagens
Modifications on lysine 87 from both ␣-chains of type V collagen in WT, Sc65 Ϫ/Ϫ , and P3h3 Ϫ/Ϫ mice were measured using mass spectrometry in skin and bone. Lysine modifications include unmodified Lys, hydroxylysine (Hyl), galactosyl-hydroxylysine (G-Hyl), and glucosylgalactosyl-hydroxylysine (GG-Hyl). Each percentage was determined based on the ratio of the m/z peaks of each post-translational variant from a minimum of three mice unless indicated. sylated aldimine cross-link present in WT tissue (Fig. 8). Such an abnormal cross-link arrangement for skin can explain the marked increase in ␥ 112 trimer and decrease in ␤ 12 dimer seen on SDS-PAGE (Figs. 2 and 3). It can also explain the increased extractability of type I collagen in non-denaturing 3% acetic acid at 4°C (ϳ40% more extractable in P3h3 Ϫ/Ϫ skin and tendon), if in essence intermolecular C-telopeptide aldols are replaced by intramolecular aldols. Although beyond the scope of the present study it should be noted that major complex cross-links on the allysine aldehyde pathway were reported to be the product of an allysine aldol linked to a histidine by Michael addition and a hydroxylysine by aldimine addition (dehydrohistidinohydroxymerodesmosine (HHMD) on reduction) (36). The structure was based on analyses of acid-hydrolyzed collagen after sodium borohydride reduction, so the native state of the mature aldol cross-linking structures in tissue remains uncertain (37). The reason why skin appears to be more affected than tendon collagen is probably related to the lack of glcgal-Hyl at Lys-87 in normal mature tendon type I collagen. To speculate, the combined loss of hydroxylation and glycosylation at Lys-87 of skin, aorta, and cornea may have the above summarized effect on C-telopeptide aldol placement but not in  tendon, which normally has Hyl alone at Lys-87 with no attached sugars. It is notable that the band pattern of WT tendon collagen on SDS-PAGE (Fig. 2) resembles that of Sc65 Ϫ/Ϫ and P3h3 Ϫ/Ϫ skin with a prominent ␥ 112 trimer and with no obvious difference from Sc65 Ϫ/Ϫ and P3h3 Ϫ/Ϫ tendons. In a sense, therefore, the effect in null mice on skin collagen is to make it resemble normal tendon collagen at this cross-link site. However, with that in mind, normal tendon collagen in a study of fibromodulin-null mice was shown to have incomplete C-telopeptide lysine conversion to aldehydes, a restriction that was lifted in the fibromodulin-null mice (22). It was speculated that in fibromodulin-null mice an increase in collagen molecules with two C-telopeptide allysines produced an observed increase in C-telopeptide aldols and ␥ 112 trimers (22).

Modification
The observed effects on collagen lysine hydroxylation and cross-linking closely resemble those observed in tissues of EDS VIA patients. Of the many defined clinical subtypes of EDS, most are characterized by skin hyper-extensibility, joint hypermobility, and connective tissue fragility. The kyphoscoliotic type, EDS VIA, is caused by homozygous or compound heterozygous mutations in the PLOD1 gene (encodes LH1). Clinical features of patients with this EDS subtype include progressive kyphoscoliosis, soft velvety skin, muscle hypotonia, arterial ruptures, and fragility of ocular tissues often leading to rupture of the globe and retinal detachment (38,39). Similar to Sc65 Ϫ/Ϫ and P3h3 Ϫ/Ϫ mice, the LH1 knock-out (Plod1 Ϫ/Ϫ ) mouse seemed to have milder consequences than Plod1 human cases. For example, muscle hypotonia and aortic ruptures were reported in the Plod1 Ϫ/Ϫ mouse but not kyphoscoliosis or skin laxity (40). Despite Sc65 Ϫ/Ϫ and P3h3 Ϫ/Ϫ mice having at this stage primarily a biochemical phenotype, the potential is clear for suppressed lysine hydroxylation to manifest as an EDS variant in humans. In the fibril, collagen molecules are spatially arranged such that intermolecular cross-link placement is optimal (B). In WT skin, fully glycosylated Hyl87 preferentially forms an intermolecular aldimine cross-link with a C-telopeptide lysine aldehyde (C). In the P3h3 Ϫ/Ϫ and Sc65 Ϫ/Ϫ mouse tissues the LH1 substrates are under-hydroxylated and subsequently under-glycosylated, which alters collagen cross-linking chemistry (D). From mutant skin the results are consistent with the C-telopeptide lysine aldehydes preferentially forming intramolecular aldol cross-links (as opposed to intermolecular aldols). We predict that under normal conditions the presence of the disaccharide on Hyl87 favors aldimine formation with a single C-telopeptide aldehyde and hinders intramolecular aldol formation with a second ␣1(I) C-telopeptide from the same molecule as the first one, so favoring intermolecular interactions. The net effect of under-hydroxylated Lys-87 then would be fewer stable aldol intermolecular cross-links within and between fibrils.
Tissues from two strains of Sc65 Ϫ/Ϫ mice (19,21) were analyzed in comparison to P3h3 Ϫ/Ϫ and respective WT tissues. Briefly, for the initial strain the embryonic stem cell clone (129X1/SvJ ϫ 129S1) with a proviral insertion (gene-trap) in the Sc65 gene was purchased from the Gene Trap Resource of the Phenogenomics Toronto Center (Toronto, Canada). C57/ BL6J blastocysts were injected with ES cells and chimeric mice were generated (19). The second Sc65 Ϫ/Ϫ strain employed a conditional targeting construct (using the vector p-flrt-neo-2XloxP) that was designed to introduce loxP sites flanking the last 2 exons (7 and 8) of Sc65 (floxed mouse). Chimeric mice were generated by electroporating the construct into mouse ES cells derived from F1(C57B6/jx129sv) embryos. Global Sc65 excision was achieved by mating male Sc65 mice carrying a floxed allele with Hprt-Cre females (21). The use of the latter laboratory mice was approved by the University of Arkansas for Medical Sciences IACUC committee. All mice used in this study had the same genetic background. The reported posttranslational variances did not appear to be related to the age or sex of the animals tested.
Collagen Extraction-Type I collagen was solubilized from bone, skin, tendon, aorta, sclera, and cornea by heat denaturation for 5 min at 100°C in Laemmli buffer (SDS extraction), 3% acetic acid at 4°C for 24 h, or cyanogen bromide digestion in 70% formic acid at room temperature for 24 h. Type II collagen (rib cage cartilage), type IV collagen (kidney cortex), and type V collagen (bone and skin) ␣-chains were solubilized by pepsin digestion or cyanogen bromide digestion as previously described (17,41). Collagen ␣-chains were resolved by SDS-PAGE and stained with Coomassie Blue R-250. Collagen extractability was determined using NIH ImageJ software as previously described (22).
P3h3 Immunoblot-Human fetal control cartilage tissue was obtained from a National Institutes of Health-sponsored institutional resource (Birth Defects Research Laboratory, University of Washington) (42). P3h3 Ϫ/Ϫ and control whole mouse kidneys were homogenized in saline with protease inhibitors (50 mM Tris-HCl, pH 7.5, 0.15 M NaCl, 2 mM PMSF, 10 mM phenanthroline). An intracellular fraction was solubilized from this tissue extract in SDS sample buffer at 95°C. Human fetal control cartilage was 4 M guanidine-HCl extracted as previously described (42). Equal loads of total protein were resolved on a 7.5% SDS-PAGE and transblotted to PVDF membrane. Membranes were blocked in 5% milk, probed with P3h3 polyclonal antibody (1:500 dilution, ProteinTech, catalog number 16023-1-AP), and detected with an HRP-conjugated secondary antibody (1:5000, anti-rabbit). Blots were developed using Super-Signal West Pico Chemiluminescent Substrate (Thermo Scientific) and exposed on autoradiography film.
H&E Staining-Skins and femurs were harvested from a 2-month-old mouse and fixed in 10% buffered formalin. Femurs were decalcified with 14% EDTA for 2 weeks before paraffin embedding processing. Paraffin-embedded skin and femur were sectioned (6 micrometer) and the sections were used for H&E staining.
Mass Spectral Analysis of Site-specific Post-translational Modifications-The post-translational modifications 3-hydroxyproline, 5-hydroxylysine (Hyl), glcgal-Hyl, and gal-Hyl glycosides were quantified at specific sites in collagen ␣-chains as previously described (41). Collagen ␣-chains were cut from SDS-PAGE gels and subjected to in-gel trypsin digestion. Skin was also digested with bacterial collagenase with and without borohydride reduction as previously described (21) and total collagenase digests were resolved into peptide fractions by C8 reverse-phase HPLC and individual fractions were analyzed by mass spectrometry (43). Electrospray mass spectrometry was carried out on the trypsin-and collagenase-digested peptides using an LTQ XL linear quadrapole ion-trap mass spectrometer equipped with in-line Accela 1250 liquid chromatography and automated sample injection (ThermoFisher Scientific). Proteome Discoverer software (ThermoFisher Scientific) was used for peptide identification. Tryptic peptides were also identified manually by calculating the possible MS-MS ions and matching these to the actual MS-MS spectrum using Thermo Xcalibur software. Hydroxyl differences were determined manually by averaging the full scan MS over several LC-MS minutes to include all the post-translational variations of a given peptide. Protein sequences used for MS analysis were obtained from the Ensembl genome database.
Collagen Cross-linking Analysis-The pyridinoline cross-link content of collagen in bone, dentin, cartilage, and aorta samples was determined by HPLC after acid hydrolysis in 6 M HCl for 24 h at 108°C. Dried samples were dissolved in 1% (v/v) n-heptafluorobutyric acid for quantitation of lysyl pyridinoline (LP) and hydroxylysyl pyridinoline (HP) by reverse-phase HPLC and fluorescence monitoring as previously described (7).