The Skeletal Structure of Insulin-Like Growth Factor I-Deficient Mice



The importance of insulin-like growth factor I (IGF-I) for growth is well established. However, the lack of IGF-I on the skeleton has not been examined thoroughly. Therefore, we analyzed the structural properties of bone from mice rendered IGF-I deficient by homologous recombination (knockout [k/o]) using histomorphometry, peripheral quantitative computerized tomography (pQCT), and microcomputerized tomography (μCT). The k/o mice were 24% the size of their wild-type littermates at the time of study (4 months). The k/o tibias were 28% and L1 vertebrae were 26% the size of wild-type bones. Bone formation rates (BFR) of k/o tibias were 27% that of the wild-type littermates. The k/o bones responded normally to growth hormone (GH; 1.7-fold increase) and supranormally to IGF-I (5.2-fold increase) with respect to BFR. Cortical thickness of the proximal tibia was reduced 17% in the k/o mouse. However, trabecular bone volume (bone volume/total volume [BV/TV]) was increased 23% (male mice) and 88% (female mice) in the k/o mice compared with wild-type controls as a result of increased connectivity, increased number, and decreased spacing of the trabeculae. These changes were either less or not found in L1. Thus, lack of IGF-I leads to the development of a bone structure, which, although smaller, appears more compact.


THE INSULIN-LIKE growth factors I and II (IGF-I and IGF-II) play important roles in the development and growth of bone. Bone cells and chondrocytes produce these growth factors,1-4) have receptors for them,5-9) and respond to them with a change in proliferation and function (e.g., collagen and glycosaminoglycan production, respectively).1, 10-17) Although the relative concentrations of IGF-I and IGF-II in bone differ among species and change from the embryonic period to the postnatal period(9, 12) both appear to be necessary for normal development as indicated by mouse models in which the genes for either growth factor led to a reduction in size at birth.18-20) Deletion of the IGF-I receptor, which mediates the actions of both IGF-I and IGF-II, leads to a more severe phenotype than the deletion of either IGF-I or IGF-II alone.(19, 21) However, postnatal development at least in the mouse seems more dependent on IGF-I than IGF-II in that the IGF-I-deficient mouse has a higher perinatal mortality and more severe postnatal growth retardation than the IGF-II-deficient mouse.18-21) Part of this difference may lie in the role that IGF-I plays in the mechanism by which a number of hormones act on bone. Growth hormone (GH) is a well-known stimulus of IGF-I production in a variety of tissues including bone,(22, 23) and although GH has its own effects on bone, much of the anabolic action of GH on bone is mediated through IGF-I.(14, 15, 22, 24) Parathyroid hormone (PTH) also increases IGF-I production,(3, 16, 25, 26) and antibodies to IGF-I can block at least some of the actions of PTH on bone.(26) Similarly, other hormones with effects on bone such as cortisol,(27) thyroid hormone,(28) estrogen,(29, 30) and androgens(31) alter IGF-I levels in bone or bone cells in a manner consistent with IGF-I playing a role in the actions of these hormones on bone.

Although IGF-I has been identified clearly as an important growth factor for the skeleton, the lack of IGF-I on the skeletal system has not been evaluated in detail. In this study, we examined the tibias and L1 vertebrae from mouse deficient in IGF-I (gene knockout [k/o]) in comparison with their normal and heterozygous littermates. Although the bones of the IGF-I-deficient mice were substantially smaller than their normal littermates (NLMs), periosteal bone formation rates (BFRs) and cortical thickness were appropriate for size, and trabecular number (Tb.N) and connectivity at least in the tibias were increased substantially. These data indicate that IGF-I is not required to form a well-developed bone but may play a role in its remodeling.



IGF-I-deficient mice were developed as previously described(20) and then backcrossed into a CD-1 genetic background. The construct used for the homologous recombination contained a neomycin cassette inserted into exon 3 of the IGF-I gene. Genotyping of the offspring used polymerase chain reaction (PCR) of tail DNA with the appropriate primers to identify either the intact exon 3 or the neomycin insertion as described.(20) For experiment A the mice were raised for 3 months on standard mouse chow. To the extent possible, each IGF-I-deficient mouse was matched to two wild-type and two heterozygous littermates of the same sex. In this experiment we report the results from 34 animals divided as follows: (1) male animals—six wild-type, six heterozygotes, and three IGF-I-deficient; (2) female animals—seven wild-type, eight heterozygotes, and four IGF-I-deficient. In experiment B, focusing on the skeletal response of these mice to IGF-I and GH administration, we evaluated 32 6-week-old mice (14 IGF-I-deficient and 18 wild-type) but did not separate these groups by gender. For the latter experiment, the mice were injected subcutaneously daily for 1 week with GH, 10 μg/kg; IGF-I, 10 μg/kg; both; or vehicle. In both experiments, the mice were injected subcutaneously with demeclocycline followed in 12 days (experiment A) or 5 days (experiment B) by calcein (each 15 mg/kg) to label the mineralization fronts of the bone. Two days later, the mice were killed. In experiment A the animals were weighed; blood was obtained for IGF-I determinations; the right tibias and L1 vertebrae were obtained for fat-free weight, peripheral quantitative computerized tomography (pQCT), and microcomputerized tomography (μCT); and the left tibias were obtained for mineral apposition rate (MAR) and BFR rate. In experiment B, the analysis was limited to MAR and BFR in the tibia. These studies were approved by the Animal Use Committee of the San Francisco Veterans Affairs Medical Center where the animals were raised and studied.

Fat-free weight of bone

The tibias and L1 vertebrae were cleared of adherent soft tissue and extracted in ethanol and diethyl ether using a Soxhlet apparatus (Fisher Scientific, Pittsburgh, PA, USA). The bone was then dried at 100°C and weighed.

Bone histomorphometry

MARs and BFRs at the tibiofibular junction were performed as follows. Diaphyseal segments of the tibias were dehydrated, defatted in acetone followed by ether, and then embedded in bioplastic (Tap Plastics, Dublin, CA, USA). After polymerizing overnight, the blocks were sectioned at a thickness of 60 μm using a Leica SP 1600 circular bone saw (Leica, San Jose, CA, USA). The section containing the tibiofibular junction was digitized with a Hamamatsu video camera (Hamamatsu, Tokyo, Japan) coupled to a Leica DMR microscope), and periosteal MAR and BFR were determined using the NIH Image program (NIH Image, Bethesda, MD, USA).


Tomographic measurements of trabecular, cortical, and total bone density were performed using a Stratec XCT-960A from Norland Corp. (Fort Atkinson, WI, USA) according to previously described methods.(32) One-millimeter-thick sections were obtained throughout the length of the tibia beginning just distal to the growth plate. To compensate for the differences in bone size between the IGF-I-deficient animals and their wild-type and heterozygous littermates, the data from the first section below the growth plate is reported as representing the proximal tibia and the data from the next to the last slice is reported as representing the distal tibia.


The tibias were analyzed with a Scanco Medical μCT apparatus (Scanco Medical, Zurich, Switzerland).(33) Three-dimensional (3D) information was obtained by stacking successively measured slices on top of each other. The voxel size was 9 μm in all three spatial dimensions. One hundred twenty-eight slices were measured in each sample, covering a total of 1.15 mm of the metaphysis. To analyze the trabecular part of the tibias, the compact part of the bone was masked out in the following way. With a 3D box-shaped low-pass filter applied to the original gray-scale CT images, an artificial partial volume effect was created, which blurs out the individual trabeculae but leaves the dense compact shell intact. Then, the cortex mask is extracted with a simple thresholding operation with a fixed threshold of 20.0% of the maximal gray-scale value. To analyze the cortex, a low-pass Gaussian filter (s = 3.0) was applied to the original images, with a threshold of 10.0%. The axial position and extent are identical with the trabecular volume. The trabecular region is masked out, and a 3D component labeling of the cortex is additionally performed to extract only the main connected component. Then, the cortex is evaluated with the direct distance transformation method to calculate its thickness. Figure 1 illustrates the ability of μCT to separate the cortex from the cancellous bone with this approach.

Figure FIG. 1.

The exploded view of the proximal tibia of a mouse analyzed by μCT. The figure illustrates the separation of the cortical bone and trabecular bone that can be achieved with this technology. The analysis is performed on each region independently.

Bone volume (BV) and bone surface (BS) are calculated using a tetrahedron meshing technique generated with the Marching Cubes method.(34) Total volume (TV) is calculated from the volume of the conforming volume of interest (VOI). Mean Tb.N, mean trabecular thickness (Tb.Th), and mean trabecular separation are calculated using newly developed direct techniques based on the distance transformation of the binary object.(35) An estimation of the plate-rod characteristic of the structure is achieved using the Structure Model Index (SMI).(36, 37) For an ideal plate and rod structure, the SMI value is 0 and 3, respectively. For a structure with both plates and rods of equal thickness, the value is between 0 and 3, depending on the volume ratio between rods to plates. The geometrical degree of anisotropy (DA) usually is defined as the ratio between the maximal and the minimal radius of the mean intercept length (MIL) ellipsoid.(38, 39) However, for our irregular volumes we did not use test lines through the volume but calculated a direction distribution of the projected triangulated surfaces by calculating the scalar product of the area-weighted normal vector of each surface triangle with the discrete directions of the direction distribution. This distribution corresponds to the inverse of the usually computed intercept length distribution. The “quasi-MIL” ellipsoid then is calculated by fitting the inverse directional projected surface distribution to an ellipsoid using a least square fit. This avoids sampling problems inherent in the MIL method.(40) Connectivity density (Conn.dens) was calculated with the Euler method of Odgaard and Gundersen.(41)

Serum levels of IGF-I

IGF-I in serum was measured as previously described.(42) Acid ethanol extraction was used to remove the IGF-binding proteins (IGFBPs), and the extracted samples were assayed for IGF-I with a radioimmunoassay (RIA) kit from Nichols Institute (San Juan Capistrano, CA, USA). The detection limit is this assay was 22 ng/ml.


The data were analyzed by multivariate analysis of variance (ANOVA) and Student's t-test as appropriate using programs in Sigma-Stat (Jandel Scientific, San Rafael, CA, USA). A value of p < 0.05 was considered significant.


IGF-I is required for growth: interaction with gender

Table 1 contains the body weights of the mice at the time of death. Wild-type male mice were 28% larger than wild-type female mice. This difference was less (19%) between the heterozygous groups and not seen among the IGF-I-deficient mice. Wild-type male mice were 24% larger than heterozygous male mice and nearly 5-fold larger than the IGF-I-deficient male mice. The differences among the female mice were similar but of lesser magnitude. Both gender (p = 0.0134) and genotype (p < 0.0001) effects were significant as was the interaction between gender and genotype (p = 0.04). The serum levels of IGF-I are shown in Table 2. The IGF-I levels in the IGF-I-deficient mice were at or below the level of detection (22 ng/ml). The IGF-I levels were 35% higher in the wild-type animals than in the heterozygotes (237 ng/ml and 176 ng/ml, respectively). No gender differences were observed in the serum concentrations of IGF-I.

Table Table 1.. Body Weights of Wild-Type, Heterozygous, and IGF-I-Deficient (k/o) Mice
original image
Table Table 2.. Serum Levels of IGF-I in Wild-Type, Heterozygous, and IGF-I-Deficient Mice
original image

Bone mass is reduced in IGF-I-deficient mice consistent with reduced growth

The fat-free weights of the tibias and L1 vertebrae are shown in Table 3. As expected, the bones from the IGF-I-deficient animals were significantly smaller than those of the wild-type animals, being 28% of the normal tibia and 26% of the normal L1 vertebra. Likewise, the bones from the heterozygotes were significantly smaller than those of the wild-type animals, being approximately 92% of normal tibial size and 84% of normal vertebral size. Despite the differences in body weight between male and female mice, no gender differences in total bone weights were observed. As will be discussed in the section dealing with the μCT results, this may be because of at least in part a gender difference in cortical versus trabecular bone distribution.

Table Table 3.. Fat-Free Weights of Tibia and L1 Vertebra in Wild-Type, Heterozygous, and IGF-I-Deficient Mice
original image

BFRs adjusted for bone size are normal in the IGF-I-deficient mouse

Periosteal MARs and BFRs, respectively, were measured at the tibiofibular junction (Table 4). The IGF-I-deficient mice had an MAR (linear measurement) approximately 50% that of the wild-type mice and a BFR (areal measurement) approximately 25% that of the wild-type mice. However, when one normalizes the BFR to the total cortical area (Table 4), these differences disappear, suggesting that bone formation in the IGF-I-deficient mouse is appropriate for bone size. No significant differences were found between genders or between wild-type and heterozygous mice for these measurements.

Table Table 4.. Tibial MAR and BFR of the Tibiofibular Junction in Wild-Type, Heterozygous, and IGF-I-Deficient Mice
original image

Bone formation is increased normally by GH and supranormally by IGF-I in the IGF-I-deficient mouse

We tested whether the administration of IGF-I and/or GH to these mice would increase BFR. The response of IGF-I-deficient mice to these hormones was compared with that of their wild-type littermates. These results are shown in Table 5. Note that the MAR and BFR of the vehicle-treated 6-week animals (Table 5) are approximately 10-fold greater than those of the 4-month animals shown in Table 4, but the differences in MAR and BFR between the IGF-I-deficient animals and their wild-type littermates are of similar proportions. GH stimulated MAR and BFR approximately 70% in both IGF-I-deficient and wild-type mice, although the absolute magnitude of response was higher in the wild-type animals. IGF-I, on the other hand, increased MAR 4.6-fold and BFR 5.2-fold in the IGF-I-deficient animals but only by 43% and 46%, respectively, in the wild-type animals. The combination of IGF-I and GH increased MAR and BFR 5.0- and 5.9-fold, respectively, in the IGF-I mice but only 2- and 1.9-fold, respectively, in the wild-type mice. These results suggest that IGF-I is not essential for GH stimulation of periosteal bone formation and that the absence of IGF-I increases the sensitivity of bone to IGF-I.

Table Table 5.. Effects of GH and IGF-I on Periosteal Bone at the Tibiofibular Junction of IGF-I-Deficient Mice
original image

Total bone mineral but not bone density is reduced in IGF-I-deficient mice

Mineral and density measurements of the tibia were made by pQCT. In Table 6 the data are summarized from 1-mm sections taken just distal to the growth plate (proximal) and the penultimate section from the distal tibia (distal). Resolution of cortical and trabecular bone in these samples was considered inadequate to provide reliable information because of the small size of the bones, so that only total mineral and density measurements in these sections were tabulated. Total mineral in both the proximal and the distal tibia was significantly affected by genotype with the IGF-I-deficient animals having reduced mineral content compared with wild-type and heterozygote animals, consistent with their smaller bones. A significant difference was also seen between wild-type and heterozygous bones in the mineral content of the distal tibia, but this difference did not reach significance in the proximal tibia. In contrast to these differences in bone mineral content, bone density was not reduced in the IGF-I-deficient mice at either the proximal or distal ends of the tibia when compared with wild-type or heterozygous littermates.

Table Table 6.. Mineral and Density Measurements of the Tibia in Wild-Type, Heterozygous, and IGF-I-Deficient Mice
original image

Site-specific increases in trabecular BV (BV/TV) in IGF-I-deficient mice: interaction with gender

Tables 7 and 8 list the structural parameters measured by μCT in the proximal tibia (Table 7) and L1 vertebra (Table 8). Figure 2 shows representative images of the metaphysis of the proximal tibia from an IGF-I-deficient and wild-type mouse. Note that for comparison purposes, the bone from the IGF-I-deficient mouse is enlarged 2× to the size of the wild-type mouse. IGF-I-deficient bones were approximately 20% of the TV of wild-type tibia and 28% of the TV of wild-type L1. However, the BV shows an important difference between proximal tibia and L1 vertebrae in that it is proportionately less reduced in the proximal tibia (24% of male wild type and 37% of female wild type) than in the L1 vertebra (16% of male wild type and 21% of female wild type). As such, the BV/TV of the IGF-I-deficient tibia is significantly higher than that of the wild-type tibia in the proximal tibia but not in L1. On the other hand, cortical thickness was reduced approximately 18% in the proximal tibia and 25% in L1 of the IGF-I-deficient mouse compared with wild-type mouse. Thus, the surprising increase in BV/TV in the IGF-I-deficient mouse was not accompanied by a similar increase in cortical thickness.

Figure FIG. 2.

Images of the metaphyses of the proximal tibia from a male wild-type and a male IGF-I-deficient mouse. These examples illustrate some of the essential differences including increased BV/TV, Conn.dens, Tb.N, and decreased Tb.Sp. The IGF-I-deficient bone has been magnified 2× relative to the wild-type bone.

Table Table 7.. Structural Parameters of the Proximal Tibia in Wild-Type, Heterozygous, and IGF-I-Deficient Mice
original image
Table Table 8.. Structural Parameters of L1 Vertebra in Wild-Type, Heterozygous, and IGF-I-Deficient Mice
original image

The wild-type male animals had larger proximal tibias than the female animals as indicated by the larger TV (45% increased in male animals) and BV (106% increased in male animals). BV/TV tended to be greater in the male animals but this did not reach statistical significance. These differences were less striking in the heterozygotes and IGF-I-deficient animals. L1 did not show this gender effect. Although wild-type and heterozygous male mice had greater TV, BV, and tended to have greater BV/TV in their proximal tibias than female mice, the reverse was true for cortical thickness, which was significantly higher in the proximal tibias of the female mouse. This was not true for L1. The increased cortical thickness of the female tibia may explain the surprising lack of gender difference in the fat free weight (FFW) of the tibia (Table 3) described previously.

The increase in BV/TV is caused by increased Tb.N and connectivity

The increase in BV/TV in the proximal tibia of the IGF-I-deficient mouse compared with wild-type littermates was caused primarily by a significant increase in both Tb.N (54% increase in male mice and 96% increase in female mice) and Conn.dens (126% increase in male mice and 417% increase in female mice). Tb.N. also increased significantly in the L1 vertebrae of IGF-I-deficient mice (29% increase in male mice and 20% increase in female mice) but not to the same extent as in the proximal tibia, and Conn.dens in L1 was not altered significantly by the lack of IGF-I. On the other hand, Tb.Th was modestly but significantly reduced in the IGF-I-deficient proximal tibia (23% decrease in male mice and 20% decrease in female mice) and L1 (30% decrease in male mice and 28% decrease in female mice). The combination of the greater increase in Tb.N and lesser decrease in Tb.Th in the proximal tibia compared with L1 of the IGF-I-deficient mouse resulted in a more marked decrease in trabecular spacing (Tb.Sp) in the proximal tibia (37% decrease in males, 52% decrease in females) compared with L1 (22% decrease in male mice and 18% decrease in female mice) of the IGF-I-deficient mouse compared with wild-type mice. As implied by the magnitude of the changes noted, the proximal tibias of female IGF-I-deficient mice were affected more profoundly by the lack of IGF-I than were those of male IGF-I mice with respect to Tb.N and Tb.Sp. These gender differences were not observed in L1. Furthermore, the DA declined in the proximal tibia of the IGF-I-deficient mouse with no significant change in DA of L1. On the other hand, the SMI was not significantly altered by IGF-I deficiency for the proximal tibia but did suggest a shift to a more rodlike architecture (higher SMI) in the L1 vertebrae of the IGF-I-deficient mouse.


IGF-I clearly plays an important role in somatic growth. This is seen readily in the smaller size of IGF-I-deficient pups at birth and the growth retardation postnatally.18-20) The question this study seeks to address is whether lack of IGF-I leads to a smaller version of a normal mouse skeleton or whether there are specific changes in bone structure that can be attributed to IGF-I deficiency. Our results indicate the latter, and the changes were not necessarily what one would have predicted. In particular, periosteal bone formation at the tibiofibular junction was surprisingly well preserved both in the young mice (6 weeks old) and in the older mice (4 months old) lacking IGF-I when the difference in size of bone was taken into account, and the density of trabecular bone in the proximal tibia actually was increased. Although these changes were less striking in the spine (L1), IGF-I deficiency did not lead to osteopenia as one might have predicted.

There may be several reasons why the IGF-I-deficient animals are able to develop and maintain bone, perhaps even above normal levels. First, is the observation that the IGF-I-deficient bone remains sensitive to GH with respect to periosteal bone formation. This was a surprise, because much of the effects of GH on bone are thought to be mediated through its stimulation of IGF-I production.14, 15, 21-23) However, as Ohlsson et al.(22) have summarized recently, GH appears to have its own actions on bone independent of IGF-I, a conclusion supported by the results presented in this report. Although we did not measure GH in this study, because of the ability of IGF-I to suppress GH secretion, one might expect the GH levels to be elevated in the IGF-I-deficient mouse compared with NLMs. This expectation is supported by recent observations that showed that although GH-containing cells in the pituitary of IGF-I-deficient mice were not increased in number, their content of GH messenger RNA (mRNA) per cell was 50% increased above wild-type mice.(43) Thus, increased GH in the IGF-I-deficient animals may help maintain a normal BFR. Furthermore, the bones of IGF-I-deficient animals proved hypersensitive to IGF-I, showing a 5-fold stimulation of bone formation when IGF-I was administered for 1 week. The bone formation in wild-type animals was stimulated only 1.5-fold. Although this is consistent with a selective up-regulation of the IGF-I response system in bone of the IGF-I-deficient animals, it also may represent a more generalized hyperresponsiveness to or changes in the levels of other anabolic agents that may or may not involve IGF-I in their mechanism of action on bone. This possibility remains to be tested.

A second explanation for the structural changes in the bones of IGF-I-deficient mice is that the balance of remodeling between formation and resorption is altered in favor of formation. IGF-I receptors have been found in preosteoclasts,(44) and IGF-I directly or through its action on osteoblasts stimulates osteoclast formation and function.45-47) Thus, IGF-I deficiency may lead to a relative decrease in bone resorption. This altered balance may differ between trabecular and cortical bone. In rats, IGF-I consistently increases periosteal bone formation, but its effects on trabecular bone formation have been variable with increased(48) and decreased(49) trabecular bone formation being reported. Thus, IGF-I deficiency might be expected to lead to decreased cortical bone but increased trabecular bone. The findings of our study on the tibia are consistent with such regional differences. However, L1 vertebrae did not show this pattern in that BV/TV was reduced in the IGF-I-deficient vertebrae. These data suggest that not only do cortical and trabecular bone differ in their response to IGF-I (or IGF-I deficiency), but the trabecular bone of different bones respond differently. The mechanism by which different bones differ in their responsiveness to hormone/paracrine factors is not understood.

Gender affected the results in the tibia; L1 did not show gender effects. Furthermore, these gender differences varied according to the region of bone studied. For example, wild-type male mice had larger cortical areas in the shaft than did wild-type female mice; yet, the cortical thickness of the proximal tibia of female wild-type animals was greater than their male counterparts. However, male wild-type mice had greater TV, BV, BV/TV, and Tb.N. and less Tb.Sp. in the proximal tibia than did female wild-type mice. These gender differences were influenced by genotype such that the differences observed between genders in the wild-type mice generally were less in the heterozygotes and not found in the IGF-I-deficient mice. These data indicate that the actions of IGF-I on bone are sexually dimorphic and suggest an interaction between sex steroid hormones and IGF-I in these actions. A similar situation appears in humans in whom male patients with GH deficiency secondary to adult-onset hypopituitarism were found to be more responsive to GH with respect to bone metabolism and increases in bone mass than were female patients.(50) One explanation for these data is that estrogen interferes with IGF-I production,(29, 30) increases the levels of the inhibitory IGFBP-4,(50, 51) and reduces circulating IGF-I levels.(52) A second explanation is that androgens may promote IGF-I production, IGF-I receptor concentrations, as well as reducing the concentrations of IGFBP-4.(31, 53) However, these observations have not been made in mice and so may not be directly applicable to this study. Nevertheless, our data are consistent with a role for sex hormones in regulating the effects of IGF-I on bone.

The structural differences discussed previously were found using μCT. pQCT failed to show an increase in the density of the proximal tibia in the IGF-I-deficient mouse. Because of the difficulty in resolving trabecular and cortical bone with this method in such small animals, only total density was recorded. Thus, the thinner cortical thickness of the proximal tibia of the IGF-I-deficient mice may have obscured the increased amount of trabecular bone (BV/TV) in this region in these animals when measured by pQCT.

In summary, the IGF-I-deficient mice show marked growth retardation, and the bones of these mice are substantially smaller than those of their wild-type littermates. However, periosteal bone formation is well maintained when the smaller size of bone is factored in possibly because it remains responsive to GH. The proportion of trabecular bone in the proximal tibia of the IGF-I-deficient mouse actually is greater than that in the wild-type mouse, although this is not observed in L1. IGF-I also affects the sexual dimorphism in tibial structure such that the gender differences observed in wild-type mice are not observed in IGF-I-deficient mice. Overall, this study indicates that the net effect of IGF-I on bone structure is complex, region and bone specific, and influenced by other hormones including sex steroids.


We acknowledge the administrative support of Victoria Lee and Vivian Wu. This work was supported by grants National Aeronautics and Space Administration (NASA) NAG2-1371, National Institutes of Health (NIH) RO1 AG13612, and NIH RO1 AR46105.