|Home | About | Journals | Submit | Contact Us | Français|
Recessive inactivating mutations in human matrix metalloproteinase 2 (MMP2, gelatinase A) are associated with syndromes that include abnormal facial appearance, short stature, and severe bone loss. Mmp2-/- mice have only mild aspects of these abnormalities, suggesting that MMP2 function is redundant during skeletal development in the mouse. Here, we report that Mmp2-/- mice with additional mutations that render type I collagen resistant to collagenase-mediated cleavage to TCA and TCB fragments (Col1a1r/r mice) have severe developmental defects resembling those observed in MMP2-null humans. Composite Mmp2-/-;Col1a1r/r mice were born in expected Mendelian ratios but were half the size of wild-type, Mmp2-/-, and Col1a1r/r mice and failed to thrive. Furthermore, composite Mmp2-/-;Col1a1r/r animals had very abnormal craniofacial features with shorter snouts, bulging skulls, incompletely developed calvarial bones and unclosed cranial sutures. In addition, trabecular bone mass was reduced concomitant with increased numbers of bone-resorbing osteoclasts and osteopenia. In vitro, MMP2 had a unique ability among the collagenolytic MMPs to degrade mutant collagen, offering a possible explanation for the genetic interaction between Mmp2 and Col1a1r. Thus, because mutations in the type I collagen gene alter the phenotype of mice with null mutations in Mmp2, we conclude that type I collagen is an important modifier gene for Mmp2.
Remodeling of the extracellular matrix (ECM) by matrix metalloproteinases (MMPs) is critical for development of the skeleton, and mice deficient for Mmp9, Mmp13, and Mmp14 (Mt1-mmp) have skeletal defects (Vu et al., 1998; Holmbeck et al., 1999; Zhou et al., 2000; Stickens et al., 2004). Inactivating mutations in human MMP2 are the cause of rare but severe skeletal defect syndromes in patients born from consanguineous marriages (Martignetti et al., 2001; Zankl et al., 2005; Rouzier et al., 2006). The clinical findings in these MMP2-null patients include hyperextension of metacarpophalangeal joints, flexion contractures of large joints, dysmorphic faces, significant growth restrictions and, surprisingly, loss of bone mass with osteolysis (Al Aqeel et al., 2000; Al-Mayouf et al., 2000; Zankl et al., 2005; Rouzier et al., 2006). In contrast to MMP2-null humans, Mmp2-/- mice have mild skeletal defects and no focal osteolysis (Inoue et al., 2006).
Genetic modifiers are genes that interact with a phenotype-associated gene and alter the observed phenotype, for example the severity of the phenotype (Nadeau, 2001). Genetic modifier studies in mice can be powerful tools to identify genetic interactions with relevance for human disease. A recent example comes from cystic fibrosis, one of the most common human autosomal recessive diseases. Mutations in cystic fibrosis transmembrane conductance regulator (CFTR), which encodes a membrane-bound chloride ion channel, cause impaired fluid and salt secretion resulting in mucus accumulation in various tissues. The clinical presentation of cystic fibrosis can vary with age-of-onset and disease severity and this variability has been ascribed both to differences in the specific mutations in the CFTR gene and to genetic modifiers (Nadeau, 2001). Mice deficient for the Cftr gene (Cftr-/- mice) die soon after birth on most genetic backgrounds, but live for many months on others. These observations indicate the existence of a modifier gene that modulates the severity of the phenotype of Cftr-/- mice. Through the intercrossing of mouse strains with different phenotypes and subsequent linkage analysis, a single locus with a semidominant effect on viability was identified (Rozmahel et al., 1996). The identification of the modifier locus in the mouse led to identification of the corresponding human locus, which turned out to be a genetic modifier of the human disease cystic fibrosis (Zielenski et al., 1999).
In the case of MMP2, the difference between the phenotype of Mmp2-/- mice and MMP2-null humans has been attributed to redundancy of Mmp2 gene function in the mouse (Martignetti et al., 2001). Because Mmp14-/- mice have severe skeletal defects with craniofacial abnormalities, growth restrictions, and osteopenia (Holmbeck et al., 1999; Zhou et al., 2000), it has been suggested that MMP14 in the mouse serves a function similar to MMP2 in human skeletal development (Martignetti et al., 2001). However, of the MMP2-null patients with osteolytic syndromes, all but one have been of Arabic descent and all have been born from consanguineous marriages. Thus, it is possible that there are genetic modifiers of MMP2, which have cosegregated with the MMP2 mutations in the affected individuals, and that such genetic modifiers are responsible for the reported different phenotypes of MMP2-null humans and Mmp2-/- mice. We had noticed that a low percentage (~10%) of Mmp2-/- mice from an inbred line on the C57BL/6 background were severely runted, had abnormally shaped faces, a wide gait, failed to thrive, and rarely lived beyond 4 weeks of age. The runted mice were always Mmp2-/-, never wild-type or Mmp2+/-. The low penetrance of the runted phenotype of Mmp2-/- mice indicated existence of a recessive genetic modifier of Mmp2 in the mouse. Rather than identifying the putative modifier through linkage analysis, we took a candidate approach. The small stature and abnormal craniofacial features of the runted Mmp2-/- mice resembled the appearance of Mmp14-/- mice (Holmbeck et al., 1999; Zhou et al., 2000), but because Mmp2-/-;Mmp14-/- double-deficient mice die immediately after birth (Oh et al., 2004), the spontaneous mutation was unlikely to directly involve Mmp14. MMP14 activates proMMP2 by cleaving off its prodomain (Strongin et al., 1995); thus, we speculated that a genetic modifier of Mmp2 could be another substrate of MMP14, namely type I collagen, which is also abundant in bones (Holmbeck et al., 1999).
Type I collagen is a heterotrimer molecule composed of two α1(I) and one α2(I) polypeptide chains encoded by two separate genes: Col1a1 and Col1a2 (Cutroneo, 2003). Type I collagen has both structural and signaling functions, the latter being mediated by integrins and discoidin domain receptors (Shrivastava et al., 1997; Heino, 2000). Type I collagen is cleaved into characteristic TCA and TCB fragments at approximately one quarter and three quarters of the length of the native molecule by collagenolytic MMPs, including MMP1, MMP2, MMP8, MMP13, and MMP14 (Ohuchi et al., 1997). These cleavage fragments are present in bone (Holmbeck et al., 1999; Stickens et al., 2004) and can initiate signaling that affects cell migration and survival (Montgomery et al., 1994; Petitclerc et al., 1999; Stringa et al., 2000; Fera et al., 2004).
Type I collagen is abundant in the skeleton where it contributes to load-bearing capacity by assembling into ordered fibrils. Naturally occurring mutations in type I collagen genes render type I collagen molecules more fragile and lead to osteogenesis imperfecta, characterized by profound skeletal fragility (Barsh et al., 1982, 1985; Bonadio et al., 1985). Conversely, Col1a1r/r mice with homozygous, engineered mutations (r) in the Col1a1 gene, which render type I collagen resistant to cleavage to TCA and TCB fragments by collagenolytic MMPs (including MMP2, MMP8, MMP13, and MMP14), continually deposit new bone, resulting in progressive thickening of the calvariae and increased trabecular bone density (Liu et al., 1995; Zhao et al., 2000; Chiusaroli et al., 2003). Mice deficient for MMP13 or MMP14, which both have high activity against type I collagen, also display abnormal skeletal development (Holmbeck et al., 1999; Zhou et al., 2000; Stickens et al., 2004).
Here, we test the hypothesis that type I collagen is a modifier of Mmp2 during skeletal development in the mouse. We report that mice that concomitantly are deficient for Mmp2 and carry the mutant Col1a1r gene show skeletal changes resembling those reported in the human skeletal syndromes associated with inactivating mutations in MMP2.
Col1a1r/r mice have mutations in the Col1a1 gene that render type I collagen resistant to cleavage by collagenolytic MMPs (Liu et al., 1995). To determine whether type I collagen was a modifier of Mmp2, we intercrossed Mmp2-/- mice with Col1a1r/r mice. When the resulting Mmp2+/-;Col1a1r/+ mice were further intercrossed, Mendelian ratios of the nine possible genotypes were observed during embryogenesis and immediately after birth (Table 1). However, at weaning, 21 days after birth, the distribution of the genotypes was significantly different from the expected values with fewer than expected Mmp2-/-;Col1a1r/+ and Mmp2-/-;Col1a1r/r mice (Table 1, χ2 = 24.7; P = 0.002). Thus, whereas Mmp2-/- and Col1a1r/r mice had normal survival, Mmp2-/- mice with at least one Col1a1r allele showed reduced postnatal survival.
The Mmp2-/-;Col1a1r/r mice were only approximately half the size of wild-type, Mmp2-/- or Col1a1r/r mice (Fig. 1A). A few Mmp2-/-;Col1a1r/r mice survived for 16 weeks or longer and their body weight was recorded weekly. At birth, the body weight of the Mmp2-/-;Col1a1r/r mice was comparable to wild-type, Mmp2-/-, and Col1a1r/r mice. However, the weight of the Mmp2-/-;Col1a1r/r mice failed to increase from approximately the fourth week of age, coinciding with the growth spurt of the long bones, and remained significantly less than the weights of wild-type, Mmp2-/-, or Col1a1r/r mice for the duration of their lifetime (Fig. 1B). The runted phenotype showed 100% penetrance in Mmp2-/-;Col1a1r/r mice and was also often found in Mmp2-/-;Col1a1r/+ mice, whereas Mmp2+/-;Col1a1r/r animals were not runted (data not shown). None of the Mmp2-/-;Col1a1r/r mice were able to produce offspring.
Humans with inactivating MMP2 mutations have characteristic swelling of fingers and toes, hyperextension of metacarpophalangeal joints, and either hyperextension or flexion contracture of most other joints (Al Aqeel et al., 2000). While handling Mmp2-/-;Col1a1r/r mice, we noticed that their range of motion, particularly of the knee, hip, shoulder, and elbow joints, was markedly reduced (8 of 8 mice). Strikingly, paws of the Mmp2-/-;Col1a1r/r mice became visibly swollen, and the metacarpophalangeal joints hyperextended around the time of weaning (Fig. 1C). To establish if the increased size of their paws was due to edema, we determined the apparent diffusion coefficient (ADC) using diffusion weighted magnetic resonance imaging (MRI). The ADC is an indirect measure of water content with a higher value reflecting increased water mobility (Le Bihan et al., 1986), which can result from the expanded interstitial space characteristic of edema. The ADC was greater in hind paws of Mmp2-/-;Col1a1r/r mice ([1.30 ± 0.05] × 10-3 mm2/sec, n = 2) than in control mice ([0.98 ± 0.03] × 10-3 mm2/sec, n = 2, one aged-matched wild-type and one Mmp2+/-;Col1a1r/r littermate). These results indicate that the paws of the Mmp2-/-;Col1a1r/r mice were edematous. Tissue involved in edema was most likely soft tissue because the T2-hyperintense regions with the highest ADC values were located at the periphery of the paws (data not shown).
Humans with recessive mutations in MMP2 have flattened faces (Al Aqeel et al., 2000; Al-Mayouf et al., 2000; Zankl et al., 2005). The facial features were also changed in the Mmp2-/-;Col1a1r/r mice, which had shortened snouts and bulging skulls (Fig. 2A,B). Facial changes of the Mmp2-/-;Col1a1r/r mice were accompanied by incomplete intramembranous ossification of calvarial bones (Fig. 2C,D). Even in the single 5-month-old Mmp2-/-;Col1a1r/r mouse that we were able to study, calvarial bones had not completely ossified (data not shown). Instead of bone, a sclerotic membrane was present in adult animals. Mild delays were also evident in Mmp2-/- and Col1a1r/r mice, but the cranial sutures of the single mutants were closed completely by 8 weeks of age (Fig. 2C). The cranial bones that developed in Mmp2-/-;Col1a1r/r mice had a very abnormal histology with a thin and porous appearance (Fig. 2E). The bones from Col1a1r/r mice were slightly osteopetrotic as previously reported (Zhao et al., 2000).
In contrast to the severe defects in cranial development, long bones of Mmp2-/-;Col1a1r/r appeared largely normal, although smaller, when examined by whole skeletal preparation (Fig. 3A). Humans with MMP2-null mutations suffer from osteolytic lesions in their long bones that can be so debilitating that they become wheel-chair-bound (Al-Mayouf et al., 2000). Microcomputed tomography (micro-CT) analysis of Mmp2-/-;Col1a1r/r mice showed reduced trabecular bone density in tibiae, indicating osteopenia (Fig. 3B; due to the reduced survival of the Mmp2-/-;Col1a1r/r mice, only one animal was analyzed at 16 weeks of age). The relative bone volume (BV/TV), which takes into account that Mmp2-/-;Col1a1r/r mice are significantly smaller than their controls, was much lower in tibiae from Mmp2-/-;Col1a1r/r than from wildtype, Mmp2-/-, or Col1a1r/r agedmatched controls (Fig. 3C). Similarly, trabecular thickness (DT-Tb.Th) and cortical thickness at the diaphysis (DT-C.Th-Diaph) were lower in Mmp2-/-;Col1a1r/r mice than in controls. Trabecular separation (TRI-Tb.Sp) was increased in Mmp2-/-;Col1a1r/r mice compared with controls, indicating active bone resorption (Fig. 3C); however, the increase was skewed by the one 16-week-old Mmp2-/-;Col1a1r/r mouse. Interestingly, this mouse had the lowest bone volume and highest trabecular separation of all the mice (Fig. 3C, red triangular symbols). Both type I collagen and MMP2 are expressed by osteoblasts in the trabecular bone (Dacquin et al., 2002; Stickens et al., 2004), and Safranin-O/Fast Green staining demonstrated that there was less trabecular bone in tibiae of Mmp2-/-;Col1a1r/r than in control Mmp2+/-;Col1a1r/r littermates (Fig. 3D). In addition, the number of tartrate-resistant acid phosphatase (TRAP)-positive osteoclasts was increased in both trabecular and cortical bone in Mmp2-/-;Col1a1r/r mice compared with aged-matched controls (Fig. 3E).
MMP2 is important for degradation of denatured, cleaved type I collagen, and, although it is often not considered a type I collagenase, it also has activity against native type I collagen (Tournier et al., 1994; Aimes and Quigley, 1995; Tam et al., 2004). The genetic interaction between Mmp2 and Col1a1r/r indicated that type I collagen could be a direct substrate of MMP2. Therefore, we compared the ability of MMP2 to degrade normal and mutant collagen in vitro with two classic collagenolytic MMPs: MMP13 and MMP14. At 25°C, MMP13 and MMP14 degraded wild-type collagen, resulting in generation of the classic TCA fragments, MMP2 had no activity against wild-type collagen, and none of the MMPs degraded the mutant collagen (Fig. 4A), as previously reported (Liu et al., 1995). However, at the physiological temperature of 37°C, MMP2 degraded wild-type and, even more efficiently, mutant collagen, but degradation did not generate TCA fragments. In contrast, MMP13 and MMP14 still had reduced activity against mutant collagen at 37°C (Fig. 4B). Thus, the genetic interaction between Mmp2 and Col1a1r could be caused by a requirement for MMP2 activity to degrade mutant type I collagen.
Humans null for MMP2 have severe osteolytic syndromes, whereas Mmp2-/- mice have only mild skeletal defects (Itoh et al., 1997; Al Aqeel et al., 2000; Al-Mayouf et al., 2000; Martignetti et al., 2001; Zankl et al., 2005; Inoue et al., 2006; Rouzier et al., 2006). The spontaneous occurrence of a low penetrance, runted phenotype with abnormal craniofacial features in our Mmp2-/- mouse colony led us to hypothesize that there were genetic modifiers of Mmp2. Rather than identifying the specific genetic modification in our line, we took a candidate approach: proMMP2 is a substrate of MMP14, and we speculated that type I collagen, another classic MMP14 substrate with established function in bone development, could modify Mmp2. Indeed, severe skeletal developmental defects were observed when Mmp2-/- mice were intercrossed with Col1a1r/r mice, which carry targeted mutations in the Col1a1 gene rendering type I collagen resistant to classic collagenase cleavage. These Mmp2-/-;Col1a1r/r mice had several striking similarities to humans with MMP2-null mutations: in addition to decreased body size, these mice had osteopenic bones, craniofacial defects, failure to close calvarial sutures, decreased range of motion of joints, hyperextended metacarpophalangeal joints, and edematous paws (Table 2). Thus, we conclude that, in mice, type I collagen is a genetic modifier of MMP2, possibly because MMP2 can degrade mutant collagen (r). The Col1a1r mutation is a dominant modifier on the Mmp2-/- background, because the phenotype of Mmp2-/-;Col1a1r/+ resembled the phenotype of Mmp2-/-;Col1a1r/r mice. In contrast, the low penetrance of the spontaneously runted Mmp2-/- phenotype is suggestive of a recessive modifier. Thus, it is unlikely that the spontaneously runted phenotype is due to a mutation that renders type I collagen resistant to MMP-mediated cleavage. It, therefore, remains to be determined if the spontaneously runted Mmp2-/- mice harbor other mutations that also affect collagen metabolism or if additional, collagen-independent, genetic modifiers of Mmp2 exist.
A priori, loss of a matrix-degrading enzyme, such as MMP2, would be expected to result in reduced bone remodeling and an osteopetrotic phenotype. Therefore, it was surprising that MMP2-null mutations were associated with reduced bone mass and osteolysis in human genetic linkage studies (Martignetti et al., 2001). Our study now confirms the human findings: absence of MMP2 activity can result in osteopenia. MMP2 is a substrate of MMP14, and Mmp14-/- mice are another example of an osteopenic phenotype that results from the absence of matrix degrading activity (Holmbeck et al., 1999). Additional similarities between Mmp14-/- and Mmp2-/-;Col1a1r/r mice are the decreased body size, abnormal craniofacial features, and lack of calvarial suture closure. These similarities strongly suggest that proMMP2 and type I collagen are important substrates of MMP14 in bone development. However, MMP14 has other substrates, including proMMP13, which is important for endochondral ossification (Cowell et al., 1998; Stickens et al., 2004). Thus, a detailed comparison of endochondral ossification in Mmp14-/- mice and Mmp2-/-;Col1a1r/r mice is likely to reveal more severe defects in Mmp14-/- than in Mmp2-/-;Col1a1r/r mice. Mmp2-/-;Col1a1r/r mice developed edema of the paws, probably localized to soft tissues. In Mmp14-/- mice, soft tissue changes, arthritis, and joint destruction have been reported (Holmbeck et al., 1999). Thus, it is possible that the edema in the Mmp2-/-;Col1a1r/r paws is secondary to changes in the joints, although a partial analysis of joints did not reveal apparent destruction. As would be expected, Mmp2-/-;Mmp14-/- double null mice also exhibit skeletal defects (Oh et al., 2004); however, due to their perinatal lethality, these defects are not well characterized.
The reduced bone mass in the absence of matrix degradation may be partly explained by a reduced ability to degrade the embryonic cartilage matrix, which is required before osteoblasts can invade into the developing tissue and deposit bone matrix (Holmbeck et al., 1999). However, this finding would not explain the progressive bone loss observed in MMP2-null humans (Al Aqeel et al., 2000; Al-Mayouf et al., 2000; Rouzier et al., 2006). Because of the poor survival of the mice, we were limited to examining bone tissue from relatively young (7-8 weeks old) Mmp2-/-;Col1a1r/r mice and could not perform a proper analysis of whether the bone loss was progressive. Nevertheless, there was a strong trend for progressive bone loss by micro-CT analysis when comparing bones from 7- to 8-week-old mice with that of the one 16-week-old mouse we could analyze. Thus, our data are consistent with the clinical findings that absence of MMP2 activity results in progressive bone loss, suggesting that the balance between bone deposition and bone resorption is skewed.
It may seem particularly surprising that combining Mmp2 null mutations with the Col1a1r mutation results in such a strong osteopenic bone phenotype because Col1a1r/r mice are osteopetrotic and in response to parathyroid hormone they have a severe impairment both in the generation of osteoclasts and in bone resorption (Zhao et al., 1999; Chiusaroli et al., 2003). Intriguingly, the porous calvarial phenotype we observed in the Mmp2-/-;Col1a1r/r mice appeared as if increased bone resorption was taking place on the osteopetrotic background of the Col1a1r/r mice. Osteoclast numbers are increased in the long bones of Col1a1r/r mice even though activity is not (Chiusaroli et al., 2003). Thus, it appears that parathyroid hormone is not the right stimulus for induction of bone resorption in the Col1a1r/r mice, whereas the absence of MMP2 is.
Type I collagen regulates osteoblast and osteoclast activity in engineered mouse models (Zhao et al., 2000; Kalajzic et al., 2002), but the molecular mechanisms involved are not clear. TCA and TCB fragments of type I collagen are present in developing bone (Holmbeck et al., 1999; Stickens et al., 2004). These collagen fragments can ligate the αvβ3 integrin (Montgomery et al., 1994), and both osteoblasts and osteoclasts express αvβ3 integrins (Cheng et al., 2000; Zhao et al., 2005). Because of the Col1a1r mutation, the TCA and TCB fragments are not generated in the Col1a1r/r mice, but the mutation could also affect the stability and conformation of the collagen triple helix. Changes in stability and conformation could affect both the unfolding of the triple helix that is necessary for collagenase cleavage (Zhao et al., 2000) and the ability of the collagen molecules to interact with other ECM components. Thus, we cannot conclude whether the absence of normal or presence of mutant fibrillar collagen is important for the observed developmental defects.
The Col1a1r mutation affects the ability of all classic collagenolytic MMPs to generate TCA and TCB fragments. In vitro, MMP2 had much higher activity against mutant type I collagen than classic collagenolytic MMPs (MMP8, MMP13, and MMP14) and the strong phenotype of mice with the Col1a1r mutation in the absence of MMP2 activity indicates that other collagenolytic MMPs are unlikely to compensate for MMP2 in vivo. Therefore, the phenotype observed in Mmp2-/-;Col1a1r/r mice is likely attributable not only to a loss of MMP2 activity against type I collagen, but to a near total loss of MMP-mediated degradation of type I collagen.
MMP2 is important for degradation of denatured, cleaved, type I collagen but also has direct activity against type I collagen (Tournier et al., 1994; Aimes and Quigley, 1995; Tam et al., 2004). Interestingly, MMP2 had higher activity against mutant type I collagen than normal type I collagen at physiological temperatures, whereas the opposite was true for the classic collagenolytic MMPs. MMP2 unfolds type I collagen by a mechanism distinct from the classic collagenases (Tam et al., 2004), and might, therefore, be more efficient at degrading mutant type I collagen with slight conformational changes. We also observed reduced survival of the Mmp2-/-;Col1a1r/+ mice with one wild-type Col1a1 allele, and these mice were also often runted with stiff joints and abnormal craniofacial features. These phenotypes were not observed in the Mmp2-/+;Col1a1r/r mice with one remaining wild-type Mmp2 allele, suggesting that MMP2 activity is absolutely required for degradation of mutant type I collagen in vivo. In vitro, we demonstrated that MMP2 was able to degrade both wildtype and mutant type I collagen, and the genetic interaction between Mmp2 and Col1a1r might, therefore, be due to direct biochemical interactions between the MMP2 enzyme and one of its substrates, type I collagen.
Inactivating mutations in Mmp2 have mild effects on skeletal development in the mouse (Inoue et al., 2006). Here, we show that concomitant mutations in both Mmp2 and Col1a1 result in a severe runted and osteopenic phenotype that resembles the osteolytic syndromes found in humans with MMP2 null mutations. Our results raise the interesting possibility that genetic modifiers of MMP2 may also play a role in the human osteolytic syndromes. MMP2 was much more efficient at degrading mutant type I collagen than were the classic collagenases. Therefore, minor mutations in type I collagen or in enzymes involved in its processing, which would go largely unnoticed on their own, could result in severe disease in the absence of MMP2 activity. Of interest, approximately 35% of the population carries a polymorphism of the COL1A1 gene that increases the ratio of the type I collagen α1(I) chain to α2(I) chain and is associated with a slightly decreased bone mass (Mann et al., 2001; Ralston et al., 2006). The strength of mouse genetics has allowed us to identify a genetic interaction between Mmp2 and type I collagen in skeletal development. Similarly, mouse genetics were instrumental in the identification of a genetic modifier of the CFTR gene (reviewed in Nadeau, 2001). The human syndromes (the Winchester, Torg, and nodulosis-arthropathy-osteolysis [NAO] syndromes) associated with MMP2-null mutations vary in severity and whether or not subcutaneous fibrillar nodules are present (Zankl et al., 2007). Such variation may be accounted for if genetic modifiers of MMP2 also exist in humans. The results of our study indicate that type I collagen and enzymes involved in collagen metabolism are interesting candidate modifiers for these devastating human syndromes. In conclusion, using mouse genetics, we have confirmed the surprising results from the human genetic linkage studies: the lack of a matrix-degrading enzyme, MMP2, can result in increased bone degradation in the context of an abnormal matrix.
Mmp2-/- mice (Itoh et al., 1997) and Col1a1r/r mice (Liu et al., 1995) were bred into the FVB/n background for four or more generations before generating Mmp2-/-;Col1a1r/r mice. Mice were genotyped by polymerase chain reaction for the wild-type Mmp2 allele (primers GelA2_sense=5′-CAACGATGGAGGCACGAGTG-3′ and GelA_antis=5′-GCCGGGGAACTTGATCATGG-3′), for the Mmp2 null allele (primers GelAmut-1=5′-GACCACCAAGCGAAACAT-3′ and GelAmut-2=5′-CAAGAAGGCGATAGAAGG-3′), for the wild-type Col1a1 allele (primers Col1a1wt-1=5′-TGGACAACGTGGTGTGGTC-3′ and Col1a1-wt-2=5′-TTGAACTCAGGAATTTACCTGC-3′), and for the mutated Col1a1r allele (primers Col1a1mut-1= 5′-TGGACAACGTGGTGCCGCG-3′ and Col1a1wt-2). Mice were housed in a specific-pathogen-free environment, under light-, temperature-, and humidity-controlled conditions with food and water available ad libitum. All mice were maintained and handled according to IACUC procedures.
Mmp2+/-;Col1a1r/+ mice were interbred to determine whether all genotypes were present in the expected Mendelian ratios. Embryos were collected, and yolk sac DNA was genotyped at embryonic day 9.5 or 14.5, and pups were collected, and tail DNA genotyped at birth (p1) or at weaning (p21).
Wild-type, Mmp2-/-, Col1a1r/r, and Mmp2-/-;Col1a1r/r male and female mice were weighed weekly. At least three animals of each sex and genotype were followed until 16 weeks of age.
Whole skeleton and skull preparations were prepared by removing skin and internal organs, fixing overnight in 4% paraformaldehyde (PFA), washing in increasing concentrations of ethanol, and staining overnight with Alcian blue (Sigma A3157-10G). Skeletons and skulls were then washed in ethanol, digested in 1% trypsin in 30% sodium borate until soft tissue became transparent, rinsed, stained in Alizarin red (Sigma A5533), incubated for 18-48 hr in 0.5% KOH/0.6% H2O2, washed in increasing concentrations of glycerin, and stored in 100% glycerine.
Skulls were photographed on a stereo microscope (Leica, MZFL111) with a digital camera (Nikon, DXM1200) using ACT-1 software (Nikon). The area that had not ossified was quantified using NIH Image J 1.34s software.
Tissues were fixed in 4% PFA and decalcified in 19% ethylenediaminetetraacetic acid for 2 weeks before embedding in paraffin and sectioning at 5 μm. Deparaffinized and rehydrated sections were stained with hematoxylin and eosin using standard methods. For Safranin-O/Fast Green staining, sections were stained in Weigert’s Iron Hematoxylin (Sigma), 0.02% aqueous Fast Green (Sigma), followed by a rinse in 1% acetic acid and 0.1% aqueous Safranin-O (Sigma). Sections were reacted for TRAP activity using a leukocyte acid phosphatase kit and counterstained with 0.02% aqueous Fast Green (Sigma).
MRI images were acquired on a 1.5T Signa whole body MRI scanner (General Electric Medical Systems). Anesthetized mice were imaged in pairs (an aged-matched wild-type or littermate Mmp2+/-;Col1a1r/r mouse with an Mmp2-/-;Col1a1r/r mouse) using a wrist radiofrequency coil (Medical Advances, Milwaukee, WI) and a customized animal holder. Normal body temperature was maintained during imaging with a heating pad. Axial T2-weighted images were acquired using a two-dimensional multislice fast spin echo sequence (repetition time [TR] = 5.5 sec, echo time [TE] = 85 msec) with an 8-cm field of view (FOV), 256 × 192 matrix, and 3-mm slice thickness to obtain cross-sectional images of the hind paws of the mice. Axial diffusion weighted images were acquired for the same slice locations using a single-shot fast spin echo (TR = 9.5 sec; TE = 85 msec; FOV = 8 cm, thickness = 3 mm, 128 × 128 matrix, and b values of 0 and 600 s/mm2). Water ADC maps demonstrating water mobility in the tissues were calculated as previously described (Partridge et al., 2001), and paw ADC values were determined using regions of interest delineated on T2-weighed images.
Tibiae were collected at indicated ages and analyzed using a micro-CT system (mCT40, Scanco Medical, Bassersdorf, Switzerland). The trabeculae of the bones were scanned using a Cone-Beam type scan into 240 slices with a voxel of 7 × 7 × 7 μm. Three-dimensional structural parameters were measured directly as described (Jiang et al., 2003).
Collagens were extracted from wild-type or Col1a1r/r mouse tails, digested with APMA-activated recombinant MMPs (Calbiochem, CA) for 1 hr at the indicated temperatures, and analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis as described previously (Liu et al., 1995). Each sample contained 1 μg type I collagen.
Statistical analysis was performed using GraphPad Prism 4 and Microsoft Office Excel 2003. The distributions of genotypes after inter-breeding Mmp2-/+;Col1a1r/+ mice were compared with the expected Mendelian distributions with χ2 test (8 degrees of freedom). Growth curves of mice (repeated measurements) were analyzed by two-way analysis of variance repeated measurements test for time points with at least three mice per genotype. For males, curves were significantly influenced by genotype, P < 0.0001 (F = 173.39) and Bonferroni posttests showed that Mmp2-/-;Col1a1r/r at 4 and 6-16 weeks of age were significantly different from the wild-types (P < 0.05). For females, statistical analysis was done only on measurements from 7-16 weeks of age, because there were fewer than three Mmp2-/-;Col1a1r/r mice for weeks 1-6. The curves were significantly influenced by genotype, P < 0.0001 (F = 969.29), and Bonferroni posttests found that Mmp2-/- mice at 15 weeks of age (P < 0.05), Col1a1r/r mice at 11 weeks of age (P < 0.05), and Mmp2-/-;Col1a1r/r mice at 7-16 weeks of age were significantly different (P < 0.001) from wild-type mice. The size of the calvarial foramen, relative bone volume (BV/TV), trabecular thickness (DT-Tb.Th), cortical thickness at the diaphysis (DT-C.Th.-Diaph), and trabecular separation (TRI-Tb.Sp.) were analyzed with two-sided Student’s t-test.
The authors thank Kerstin Dehne and Jake Lee for technical support, Dr. Stephen Krane for Col1a1r/r mice, Dr. Shigeyoshi Itohara for Mmp2-/- mice, and Drs. Sylvain Provot and Andrew J. Ewald for helpful comments on the manuscript. M.E. received a fellowship from The Danish Medical Research Council, and A.E. received a fellowship from The Serono Foundation for the Advancement of Medical Science.
Grant sponsor: National Institutes of Health; Grant number: R01 AR046238; Grant number: CA094168; Grant number: CA098075; Grant sponsor: UCSF Breast Cancer SPORE P50 CA058207 and Core C of P01 CA072006; Grant sponsor: The Sandler Program in Basic Sciences; Grant sponsor: The National Technology Center for Networks and Pathways; Grant number: U54 RR020843; Grant sponsor: The Danish Medical Research Council; Grant sponsor: The Serono Foundation for the Advancement of Medical Science.