Conditional Ablation of the Osteoblast Lineage in Col2.3Δtk Transgenic Mice


  • D. Visnjic,

    1. Department of Genetics and Developmental Biology, University of Connecticut Health Center, Farmington, Connecticut, USA
    Search for more papers by this author
  • I. Kalajzic,

    1. Department of Genetics and Developmental Biology, University of Connecticut Health Center, Farmington, Connecticut, USA
    Search for more papers by this author
  • G. Gronowicz,

    1. Department of Orthopedics, University of Connecticut Health Center, Farmington, Connecticut, USA
    Search for more papers by this author
  • H. L. Aguila,

    1. Department of Medicine, University of Connecticut Health Center, Farmington, Connecticut, USA
    Search for more papers by this author
  • S. H. Clark,

    1. Department of Genetics and Developmental Biology, University of Connecticut Health Center, Farmington, Connecticut, USA
    Search for more papers by this author
  • A. C. Lichtler,

    1. Department of Genetics and Developmental Biology, University of Connecticut Health Center, Farmington, Connecticut, USA
    Search for more papers by this author
  • D. W. Rowe

    Corresponding author
    1. Department of Genetics and Developmental Biology, University of Connecticut Health Center, Farmington, Connecticut, USA
    • Address reprint requests to: David Rowe, M.D., Department of Genetics and Developmental Biology, MC 1231 (Room E-2013), University of Connecticut Health Center, 263 Farmington Avenue, Farmington, CT 06030, USA
    Search for more papers by this author


Two transgenic mouse lines were generated with a DNA construct bearing a 2.3-kilobase (kb) fragment of the rat α1 type I collagen promoter driving a truncated form of the herpes thymidine kinase gene (Col2.3Δtk). Expression of the transgene was found in osteoblasts coincident with other genetic markers of early osteoblast differentiation. Mice treated with ganciclovir (GCV) for 16 days displayed extensive destruction of the bone lining cells and decreased osteoclast number. In addition, a dramatic decrease in bone marrow elements was observed, which was more severe in the primary spongiosum and marrow adjacent to the diaphyseal endosteal bone. Immunostaining for transgene expression within the bone marrow was negative and marrow stromal cell cultures developed normally in the presence of GCV until the point of early osteoblast differentiation. Our findings suggest that the early differentiating osteoblasts are necessary for the maintenance of osteoclasts and hematopoiesis. Termination of GCV treatment produced an exaggerated response of new bone formation in cortical and trabecular bone. The Col2.3Δtk mouse should be a useful model to define the interrelation between bone and marrow elements as well as a model to analyze the molecular and cellular events associated with a defined wave of osteogenesis on termination of GCV treatment.


ALTHOUGH THE primary role of the differentiated osteoblast is the production of a bone matrix optimal for regulated mineralization, there are data to suggest that it plays a supportive role in the maintenance of the hematopoietic and immune system including the maturation and function of osteoclasts.(1) Bone marrow stromal cells (MSCs) and osteoblastic cell lines produce macrophage colony-stimulating factor (M-CSF),(2) osteoprotegerin (OPG), and OPG ligand (OPGL),(3, 4) which are regulators of osteoclast differentiation.5-7) Osteoblasts derived from human bone explants produce granulocyte colony stimulating factor (G-CSF)(8) and other factors that maintain hematopoietic progenitors and long-term culture-initiating cells (LTC-IC) in vitro.(9)

In transgenic mice expressing mutated collagen X chain, a defect in hypertrophic chondrocytes of the growth plate is accompanied by leukocyte deficiency in the bone marrow, reduction in size of the thymus and spleen, and lymphopenia.(10, 11) In mice lacking Cbfa1, there is a complete lack of osteoblast differentiation and bone formation. The membranous bones remain as a thin layer of mesenchymal tissue and the long bones persist as a cartilaginous anlage with a complete lack of marrow space and intramedullary hematopoiesis.(12, 13) These instructive in vivo models affecting skeletal development strongly support the concept that osteoblasts influence marrow function. However, these models have the disadvantage of a complex developmental history and decreased viability that complicate an analysis of the relationship of the osteoblast to other cell types. We reasoned that a developmentally normal adult mouse in which selective destruction of bone cells could be induced would be an important model to study the relationship of functional bone cells to the other cell types.

Tissue-targeted transgenic expression of herpes simplex virus thymidine kinase (HSV-tk) has been used to conditionally deplete a specific cell population.(14, 15) The enzyme is not toxic in mammalian cells but becomes toxic in the presence of nucleoside analogs like ganciclovir (GCV) or acyclovir, resulting in cell ablation. Cell specificity is determined by the promoter used to express tk. Recently, transgenic mice were generated that express the tk gene under the control of the osteocalcin promoter (OC-tk) selectively in mature osteoblasts.(16) When transgenic mice were given GCV, osteoblasts were destroyed and on removal of the drug, the bone recovered uneventfully. An unexpected finding in the analysis was the presence of active osteoclasts, suggesting that bone resorption is not coupled as tightly to bone formation as current models of bone remodeling suggest. In addition, no effect of GCV on bone marrow was observed.

We chose to drive tk in transgenic mice under control of a 2.3-kilobase (kb) type I collagen a1 (Col1a1) promoter fragment known to have restricted activity in differentiated osteoblasts but at an earlier point in their developmental lineage than OC expression. Previously, we found that newborn mice carrying a 3.6 kb rat Col1a1 promoter linked to the chloramphenicol acetyl transferase (CAT) gene showed high levels of transgene expression in bone, developing tooth germ, and tendon; lower levels in skin; and no activity in liver and brain.(17) Truncation of the 3.6-kb promoter to 2.3 kb did not affect the level of CAT activity in calvaria but caused a 2- to 4-fold drop in activity in tendon and an even greater loss in skin.(18) Subsequent work showed that a homeodomain binding sequence at −1683 base pairs (bp) is important for Col1a1 promoter activity in differentiated osteoblasts.(19) Liska et al. detected by in situ hybridization a similar cell-specific pattern of expression in transgenic mice harboring a 2.3-kb human COLIA1 promoter driving a fragment of bovine growth hormone.(20) Similarly, Rossert et al. showed a bone-restricted pattern of expression using transgenic mice carrying a murine 2.3-kb Col1a1 promoter fragment fused to either the β-galactosidase or luciferase gene.(21)

In this study, 8-week-old mice bearing the Col2.3Δtk transgene treated with GCV showed extensive ablation of all osteoblasts, a reduction in bone marrow cellularity, and a marked decrease in the number of osteoclasts. On removal of the drug, an exaggerating osteogenic response was observed. Our findings suggest that the contribution of the osteoblast to the maintenance of other cellular components of the bone marrow is dependent on the level of maturation of the osteoblast lineage.


Col2.3Δtk construction

The Δtk is a truncated version of tk gene that retains the functional properties of the enzyme without causing the sterility in transgenic males (Fig. 1A).(22) The pLTR-Δtk plasmid (a generous gift from Dr. Klatzmann(22)) lacks the internal transcription start site of the full-length tk making expression of the trans-gene dependent on the COLIA1 promoter. To construct Col2.3Δtk, the HindIII site of pLTR-Δtk was converted to BamHI and the BamHI fragment containing the Δtk gene was isolated. This fragment was cloned into BamHI site of a derivation of Cla12,(23) into which we had previously inserted a 230-bp fragment of the bovine growth hormone polyadenylation (bGH-PA) signal at the EcoRI site, and a 2.3-kb HindIII-XbaI fragment of the rat Col1a1 promoter at the HindIII-XbaI sites.(24) A 3.7-kb linear DNA fragment (Fig. 1A) containing the 2.3ColΔtk fusion gene was excised from the flanking vector sequences with ClaI and used for pronuclear microinjection of CD-1 mouse embryos to generate transgenic mice.(25)

Figure FIG. 1.

Col2.3Δtk mice. (A) Map of the Col2.3Δtk fusion gene construct. Col2.3 was derived from the rat Col1a1 gene and contains approximately 2300 bp of 5′-flanking sequence and extends into the first exon to +115 bp (Ex1). This fragment was fused to 1.1 BamHI fragment of Δtk lacking the 5′-fragment upstream of the second ATG of the parental HSV-tk gene derived from Salomon et al.(22) The arrow indicates the start of the transcription. The polyadenylation signal (pA) are provided by the bGH cassette and by HSV-tk. (B) Northern blot analysis of Δtk expression in the 98-117 and 98-200 transgenic mouse lines. Each lane contains 10 μg of total RNA prepared from different tissues of 6-week-old heterozygous male mice except for the lane marked HSV-Vero, which contains 1 μg of total RNA from Vero cells infected with herpes simplex virus and 9 μg of RNA from uninfected fibroblasts.

Administration of GCV

Δtk/+ and +/+ CD-1 mice (6 weeks of age) were injected intraperitoneally (i.p.) twice daily for 16 days with 3-8 mg/kg per day of GCV (Cytovene-IV; Roche Pharmaceutical, Nutley, NJ, USA) dissolved in physiological saline. For recovery experiments, mice were treated with GCV for 16 days and killed 4 weeks after the treatment was stopped.

RNA analysis

Animals were killed with CO2 asphyxiation and excised tissues were frozen immediately in liquid nitrogen. The epiphyseal portions of the long bones (humerus, tibia, and femur) were cleaned of attached muscle and the marrow contents were flushed using a 25G needle before freezing. Calvaria and tail containing tendon and vertebra were minced with scissors. Tail tendon samples were obtained by stripping tendon bundles away from their bony insertions. Frozen samples were homogenized in Trizol Reagent (Gibco BRL-Life Technologies, Grand Island, NY, USA) with a 5-mm Polytron (Brinkman, Westbury, NY, USA) for 30 s. Total RNA was prepared from mouse tissues and cultured cells using TRI Reagent according to the manufacturer's instructions. RNA was separated on a 2.2 M formaldehyde/1% agarose gel and transferred onto nylon membrane (Maximum Strength Nytran; Schleicher & Schuell, Keene, NH, USA). Membranes were probed with the 1.1-kb BamHI fragment of Δtk DNA, a 900-bp PstI fragment of rat Col1a1 (pα1R2),(26) a 440-bp PstI/EcoRI mouse OC fragment (p923),(27) and a 1000-bp EcoRI mouse bone sialoprotein (BSP) fragment.(28, 29)

Cell cultures

Primary calvarial cultures were prepared by sequential collagenase digestions from 6- to 8-day-old wild-type and heterozygous tk/+ neonates as previously described.(19) Cells were plated at a density of 1-1.5 × 104 cells/cm2 in six-well culture plates (Cat. No. 3046; Falcon, Franklin Lakes, NJ, USA). Twenty-four hours later the medium was changed and cells were fed again after 3 days. From day 7, the medium was supplemented with 50 μg/ml of ascorbic acid and 4 mM of β-glycerol-phosphate and was changed every other day. The cultures were tested for GCV sensitivity by addition of 2-25 μM of GCV to the media after cell attachment on the day of plating and daily thereafter. On day 7, cells in three replicate wells were trypsinized, resuspended in phosphate-buffered saline (PBS), and counted in a Coulter counter (Coulter Electronics, Ltd., Hialeah, FL, USA).

MSC cultures were prepared from bone marrow cells flushed from tibial, femoral, and humeral shafts and resuspended in α-modified essential medium (α-MEM) supplemented with 10% fetal bovine serum, 100 U/ml of penicillin, and 100 μg/ml of streptomycin. Cells were plated onto six-well plates at a density of 1.8-2 × 106 cells/cm2. Half of the media was exchanged on day 4. Beginning on day 7, the media was supplemented with ascorbic acid (50 μg/ml), β-glycerol-phosphate (8 mM), and dexamethasone (10−8 M) and was changed every 2 days. After 20 days in culture, three replicate wells were stained for alkaline phosphatase (ALP) activity (Sigma 86-R, Sigma Diagnostics Inc., St. Louis, MO, USA) and counterstained with hematoxylin or extracted for RNA.

Hematological analysis

Blood was collected in heparinized capillary tubes from the femoral vein. Hematological analysis was conducted by the Department of Laboratory Medicine at the University of Connecticut Health Center (Farmington, CN).

Bone histomorphometric analysis and immunocytochemistry

Tibias were fixed in 4% paraformaldehyde at 4°C for 7 days. After fixation, bones were decalcified in 15% EDTA for 1 week, dehydrated in progressive concentrations of ethanol, cleared in xylene, and embedded in paraffin. The entire tibia was sectioned longitudinally in 5 μm/sections. Sections from the center of the tibia were used for histomorphometry and cytochemistry. Center sections were determined by the presence of the central blood vessel and by serial sectioning to reach the widest diameter of the bone. These center sections were collected, deparaffinized with xylene, and hydrated in descending concentrations of ethanol. To determine osteoclast parameters, sections were stained for tartrate-resistant acid phosphatase (TRAP)(30) and counterstained with hematoxylin. Osteoclasts were identified by TRAP staining, their characteristic multinucleated morphology, and their location on the bone surface. Histomorphometric analysis was performed in a blinded, nonbiased manner using a BioQuant computerized image analysis system (R & M Biometrics, Nashville, TN, USA). The measurements, terminology, and units that were used for histomorphometric analysis were those recommended by the Nomenclature Committee of the American Society of Bone and Mineral Research.(31)

For immunocytochemistry, sections were placed in PBS and blocked with 3% normal goat serum in PBS. After each step, sections were washed twice with PBS. A 1:100 dilution of a rabbit anti-HSV-TK antiserum(32) in 3% normal goat serum in PBS was incubated with the sections for 2 h at 37°C. Exogenous peroxidase was blocked with 0.3% hydrogen peroxide for 30 minutes. The antibody was visualized by an immunoperoxidase, avidin/biotin method according to the Vectastain Elite ABC Kit (Vector Laboratories, Burlingame, CA, USA). Sections were counterstained with hematoxylin. Normal rabbit serum diluted 1:100 was used as a control.


Tissue expression of Col2.3ΔTK

Two transgenic founder mice, one male (line 98-117) and one female (line 98-200) carrying the Col2.3Δtk fusion gene were obtained. Both founders were fertile, transmitted the transgene at the expected frequency, and bred to homozygosity.

Northern analysis of tissues from 6-week-old Col2.3Δtk heterozygous male mice revealed a prominent 1.2-kb Δtk transcript in long bones, calvaria, and tail samples that included tail vertebrae (Fig. 1B). A much lower level of transgene expression was detected in tail tendon and brain in both lines and in aorta in line 98-117. The presence of high levels of Δtk mRNA was associated with strong Col1a1 and OC expression in samples of bone, calvaria, and tail. Low Δtk messenger RNA (mRNA) levels were associated inconsistently with a weak but detectable OC signal in isolated tendon. There were no detectable OC and Col1a1 transcripts in the sample of brain even after the film was overexposed (data not shown). There were no detectable Δtk transcripts in skin or bone marrow flushed from the long bones. A 0.9-kb Δtk transcript was detected in the testis, which represents the smallest of the multiple bands observed when full-length tk is expressed in transgenic mice.(22) Because this size is smaller than that observed in bone tissue, it is likely that the transcript is initiated from a site within the transgene as noted by others.(33, 34)

To determine the localization of the tk protein, sections from the tibias of 2-month-old male transgenic mice were stained with anti-tk antiserum (Fig. 2). Osteoblasts lining the endosteal, periosteal, and trabecular surfaces were strongly positive (Fig. 2C). Some osteocytes were stained also. High-power examination (Fig. 2D) did detect minimal staining in cells associated with the capillary wall. No staining was found in growth plate chondrocytes. Nontransgenic littermates did not show any staining (Figs. 2A and 2B). Brain, testis, ovary, and liver from transgenic mice did not show any tk staining (not shown).

Figure FIG. 2.

Immunohistochemical analysis for Δtk protein expression in tibias from a 2-month-old nontransgenic littermate at magnification (A) ×10 and (B) ×40 and from a 2-month-old Col2.3Δtk mouse at magnification (C) ×10 and (D) ×40. (A and B) No TK immunoreactivity was observed in nontransgenic tibias. (C and D)TK staining is indicated by the brown color, which is observed in osteoblasts lining the trabeculae, endosteum, and osteocytes. (A and C) Bars = 500 μm; (B and D) bars = 100 μm.

Specificity of GCV-mediated toxicity in Col2.3Δtk osteoblasts in vitro

We determined if GCV would ablate differentiated bone cells in primary calvarial or marrow stromal cultures. Neonatal calvarial cultures derived from Δtk/+ and +/+ mice were exposed continuously to control or GCV (2-25 μM) containing media 6 h after plating. After 7 days of treatment, GCV reduced the number of Δtk/+ cells by 60% at the 2-μM concentration but did not affect the number of +/+ cells (Fig. 3A). None of the GCV-resistant cells in the Δtk/+ cultures at 7 days or colonies at 14 days were ALP-positive (ALP+). After 2 weeks of treatment with 2 μM of GCV, BSP and OC expression were blocked completely whereas Δtk and Col1a1 expression were reduced dramatically in the Δtk/+ samples (data not shown).

Figure FIG. 3.

Specificity of GCV-mediated toxicity to Col2.3ΔTK osteoblasts in vitro. (A) Osteoblast number in primary calvarial cell culture derived from Δtk/+ mice grown in the absence or presence of increasing doses of GCV (2-25 μM). Osteoblast differentiation in primary MSC culture derived from Δtk/+ mice and grown in the absence or presence of 5 μM and 25 μM GCV for 20 days. (B) ALP+ colonies. (C) Northern analysis of total RNA probed for Δtk, COLIA1, BSP, and 18S ribosomal RNA (rRNA). (D) Cumulative results from three Northern experiments showing Δtk, Col1a1, and BSP mRNA levels in MSC from +/+ and Δtk/+ mice treated with 0, 5, and 25 μM GCV. Each value is the mean ± SEM.

In contrast, MSC cultures established from 6-week-old +/+ or Δtk/+ mice showed no change in the number of ALP+ colonies when grown continuously from day 1 in the presence of 5 μM of GCV (Fig. 3B). However, at the highest dose (25 μM) there was a decrease in the number of ALP+ colonies in cultures from both Δtk/+ and +/+ mice. RNA was extracted from MSC cultures treated with 5 μM and 25 μM (Fig. 3C). Cultures from +/+ mice, whether treated or not treated with GCV for 20 days, acquired the markers of differentiated osteoblasts. In the absence of GCV, cultures from Δtk/+ mice expressed markers of bone cell differentiation as well as Δtk mRNA showing that the 2.3-kb Col1a1 promoter was active. Treatment with GCV eliminated the Δtk expression and dramatically reduced Col1a1 and BSP expression. Figure 3D quantitates the intensity of the RNA transcripts relative to the 18S signal. Although there was significant variability in the degree of differentiation among the three experiments, 5 μM of GCV significantly reduced the expression of the Δtk, Col1a1, and BSP in cultures from Δtk/+ mice when compared with the untreated cultures. However, at 25 μM of GCV there was inhibition of these transcripts even in cultures derived from +/+ mice.

These results suggest that differentiated osteoblasts are the major cell population that initiates the primary calvarial cell culture and that their presence is necessary for the subsequent production of ALP+ and BSP or OC expressing colonies. In contrast, the osteoblast population in MSCs arises from an early osteoprogenitor cell that is resistant to GCV treatment and progresses to a differentiated cell that is being GCV sensitive. The residual Col1a1 expression in the GCV-treated cultures arises from the ALP+ preosteoblastic cells that do not express the Col2.3 promoter.

Tissue analysis of GCV-treated Col2.3Δtk mice

From in vitro cell culture data and in vivo expression studies, we hypothesized that transgene expression has sufficient bone specificity to allow selective ablation of osteoblastic cells without affecting other type I collagen-producing mesenchymal cells. Transgenic and control mice were treated with 3 mg/kg per day or 8 mg/kg per day of GCV for 16 days. Both control and treated mice looked healthy and had normal cage behavior. Northern blot analysis revealed the complete disappearance of the Δtk, collagen, and OC transcripts in bones, calvaria, and tail but no change in the level of Δtk transcript in the testis (Fig. 4).

Figure FIG. 4.

Northern analysis of RNA from tissues of 8-week-old male Δtk/+ mice treated with 3-8 mg/kg per day of GCV for 16 days. Total RNA (10 μg/lane) extracted from long bone, tail containing vertebral bone, and the testis was probed for Δtk, Col1a1, OC, and 18S ribosomal RNA (rRNA).

Histological analysis of tibias from nontransgenic and Δtk transgenic mice treated with GCV at a dose of 8 mg/kg per day for 16 days is shown in Fig. 5. Untreated nontransgenic and transgenic tibias appeared similar (Fig. 5A vs. Fig. 5B). Treated transgenic mice had a marked loss of bone marrow cellularity and non-transgenic-treated mice showed no evidence of GCV-induced toxicity (Fig. 5C vs. Fig. 5D). High-power examination revealed a loss of osteoblasts that normally line trabecular (Fig. 5F vs. Fig. 5E) and endosteal (Fig. 5H vs. Fig. 5G) surfaces of the treated Δtk/+ mice. TRAP cytochemistry showed significantly fewer osteoclasts in all regions of the tibia (Fig. 5F vs. Fig. 5E, see arrows, and Table 1.) Osteocytes were still present in cortical and trabecular bone of treated mice. The marrow was better preserved in the diaphyseal region although it was more severely affected adjacent to endosteal surfaces (Fig. 5H vs. Fig. 5G). The loss of cellularity within the bone marrow was confirmed by the total cell count of the cell suspension flushed from the marrow cavity of the remaining humeri and femurs of the GCV-treated Δtk/+ and +/+ mice (33.5 × 106 ±6.7 vs. 56.3 × 106 ±3.1; p < 0.05).

Figure FIG. 5.

Histology of tibias from 8-week-old +/+ and Δtk/+ mice after treatment with 8 mg/kg per day of GCV for 16 days. (A) +/+ littermates nontreated or (B) Δtk/+ nontreated, (C) +/+ treated, (D) Δtk/+ treated, (E) magnification ×20 of the metaphyseal region of tibia from treated +/+ (arrows identifies osteoclasts), (F) magnification ×20 of the metaphyseal region of the tibia from treated Δtk/+, (G) magnification ×10 of the diaphyseal region of the tibia from treated +/+, and (H) magnification ×10 of the diaphyseal region of the tibia from treated Δtk/+ mice. (A-D) Bars = 500 μm; (E-H) bars = 100 μm.

Table Table 1.. Histomorphometry of GCV-treated and GCV-recovered tk/+ and +/+ mice
original image

Histomorphometric analysis confirmed the histological findings (Table 1). The cortical width (CtWi), trabecular bone area (BAr), trabecular number (TbN), trabecular bone perimeter (BPm), and percent BAr per total tissue area (TA/TTA%) were not significantly reduced in the tibias of untreated and treated wild-type mice compared with untreated and treated Δtk/+ mice. However, osteoclast number per BAr (OcN/BAr) in the tibias of the treated Δtk/+ mice was significantly decreased relative to untreated transgenic mice and to untreated and treated wild-type mice. Histomorphometric analysis of calvaria revealed no significant reduction in total bone volume or CtWi (data not shown), but the OcN/BAr was significantly decreased (7.5 ± 4.6/mm2 vs. 1.9 ± 0.8/mm2; p < 0.001).

Peripheral blood analysis revealed that the platelet number, hemoglobin level, and mean corpuscular volume were in the normal range in all mice (data not shown). The spleen weight in transgenic mice treated with GCV was significantly (p < 0.05) increased compared with Δtk-untreated mice (110.0 ± 5.7 mg, n = 5, vs. 92.3 ± 4 mg, n = 9).

Recovery of the bones after GCV treatment

Repopulation of bone cells was examined in mice subjected to ablation of transgene-expressing cells followed by a recovery period. RNA analysis of tissues from treated mice (Figs. 6A and 6B, lanes 2-5) showed the absence of Δtk, Col1a1, and OC transcripts and a greatly reduced level of BSP in bone and calvaria. In samples derived from transgenic mice after 28 days of recovery, the level of markers of osteogenesis was increased dramatically (Figs. 6A and 6B, lanes 6-9).

Figure FIG. 6.

Northern analysis of RNA obtained from tissue of treated and recovered Δtk/+ mice. Mice were treated with 8 mg/kg per day of GCV for 16 days and then allowed to recover for 28 days. Total RNA was isolated from (A) long bones and (B) calvaria. Lane 1 represents RNA derived from Δtk/+-nontreated animals (lanes 2-5), four individual Δtk/+ mice after 16 days of treatment (lanes 6-9), four individual Δtk/+ mice 28 days after GCV was discontinued (lane 10), and nontreated +/+ mouse.

Histological examination revealed no abnormalities of the bone or the bone marrow in nontransgenic mice during the treatment or after recovery (Fig. 7A). In recovered transgenic mice, there was an excessive osteogenic response with thickened cortices and trabeculation of the marrow space in tibias (Fig. 7B). The entire medullary cavity in both the metaphysis and the diaphysis (Figs. 7C-7F) was filled with trabeculae and outgrowths of bone from the endosteal surface of cortical bone. The mineralizing trabeculae were lined with plump cuboidal osteoblasts and numerous osteoclasts (Figs. 7C and 7D, OB arrows). In many regions of the diaphysis (Figs. 7E and 7F) almost the entire medullary space was bone.

Figure FIG. 7.

Histology of tibias from (A) +/+ and (B) Δtk/+ animals treated with 8 mg/kg per day for 16 days followed by a 28-day recovery period (magnification ×4). Metaphyseal region of the Δtk/+ at (C) magnification ×20 and (D) ×40x. (E) Diaphyseal region of the Δtk/+ at magnification (E) ×20 and (F) ×40. The arrows point to numerous osteoblasts (OB) and osteoclasts (OC). (A and B) Bars = 500 μm; (C-F) bars = 100 μm.

Histomorphometric analysis of tibias from transgenic mice that had recovered for 28 days from GCV treatment showed a large increase in bone mass (Table 1). Trabecular BAr, TbN, trabecular BPm, and the BAr per TA/TTA% were significantly increased in the tibias of recovered Δtk/+ mice compared with recovered wild-type mice. OcN/BAr was now comparable with the treated, untreated, and recovered wild-type mice. Thus, in transgenic mice, the 28-day recovery from GCV treatment returned osteoclast parameters to their normal levels compared with Δtk/+ mice that were killed immediately after GCV treatment. The cortical bone width (CtWi) increased from 187 ± 9 μm in the GCV-treated transgenic mice to 339 ± 53 μm after 28 days of recovery. This increase in CtWi was not significant (p < 0.07) because of the higher SD of the GCV-treated group, in which a large increase in trabecular bone was found in the diaphysis and sometimes extending up into the metaphysis. No trabeculae were present in the diaphysis of the treated and recovered +/+ mice or the Δtk/+ treated with GCV. In all the Δtk/+ GCV-recovered mice, the medullary space of the diaphysis was filled almost entirely with bone (Fig. 7E).


The outcome of a tk cell ablation study in intact mice is highly dependent on the cell specificity of the promoter-driving expression of the tk transgene and the susceptibility of the cells to GCV toxicity. Targeted cell ablation using this strategy requires that high levels of enzyme be achieved in a subset of rapidly proliferating cells. However, the choice of a promoter that is highly tissue specific results in expression of the tk gene after full differentiation is attained at a time when the rate of cell proliferation is reduced. For example, conditional ablation of osteoblasts was achieved by using the OC promoter, which is expressed late in the osteoblast lineage after cell division has ceased.(16) For our study, we selected a bone-directed type I collagen promoter, which is activated at an earlier stage of osteoblast differentiation when cell division is thought to be still active. The results of our study indicate that a distinctly different phenotype results when ablation occurs at an early or late stage in the osteoblast lineage.

Previous work using 7-day postnatal transgenic mice showed strong activity of a rat 2.3-kb Col1a1 promoter in mature bone cells with lower expression in periosteum, tendon, skin, and brain depending on the sensitivity of the analysis.(17) This expression pattern depends in part on the presence of a homeodomain binding sequence located at −1683 bp.(19) Because the expression pattern of the 2.3-kb Col1a1 promoter had not been examined in mature mice, we determined the relative levels of the Δtk transgene in both lines by Northern analysis and immunostaining. High levels of the Δtk transcript were observed in calvariae and long bones in vivo and in osteoblastic cells in vitro. Immunostaining of tk enzyme in bone showed that the protein was localized to cells lining the trabeculae and the endosteal surfaces, which are the most active sites of bone formation. Staining did not extend into the hypertrophic cartilage cells. Much lower levels of Δtk transcript were found in isolated tendon strips (<5%) in both lines. The possibility of contamination of the sample by osteoblasts lining the tail vertebrae cannot be ruled out completely. However, the current finding of 2.3-kb promoter activity in isolated tendon is consistent with our previous work using the distal part of neonatal tails. That study examined further truncations of the Col2.3 promoter and showed that sequences downstream of −1670 bp are still capable of directing expression to tendon although the activity in neonatal calvarial bone is lost completely.(18, 35) The expression of the Δtk transgene was not detected in the sample of skin or in other type I collagen-producing tissues such as lung, bladder, or liver.

Because the transgene was expressed weakly in tissues other than bone, a series of experiments were performed to identify a dose of GCV that would selectively ablate osteoblasts without having systemic effects on the health of the mice. Between a dose range of 3 and 8 mg/kg per day, the behavior and appearance of the mice treated for 16 days were indistinguishable from controls. Under these conditions, histological analysis showed a complete loss of endosteal and trabecular lining cells in the transgenic tibias. Fewer osteoclasts were seen, which was confirmed by quantitative histomorphometry (1 ± 1/mm2 vs. 304 ± 49/mm2; Table 1). Unlike the OC-tk mice that lost bone mass because of continued osteoclast activity, the Col2.3Δtk mice maintain their bone mass during the treatment period. Northern analysis of RNA extracted from treated bone confirmed the visual picture. Expression of bone-associated transcripts Col1a1, BSP, and OC was reduced greatly and inversely proportional to the dose of GCV (Fig. 4).

The most striking histological finding of the GCV-treated bones was the marked reduction in bone marrow cellularity. The decrease in bone marrow cells was more pronounced in metaphyseal regions than in diaphyseal regions and was more prominent in areas adjacent to endosteal surfaces of the diaphyseal bone, suggesting that the changes were a consequence of the loss of osteoblasts. Previous studies on the spatial organization of hematopoietic cells in the bone marrow revealed that early hematopoietic precursors of B-cell and myeloid lineages are located preferentially in the subendosteal area, whereas their differentiated progeny migrate centrally through the sinusoids into the blood stream.(36, 37) Bone cells lining endosteal surfaces send their cytoplasmic processes into the hematopoietic spaces(38) and produce cytokines and growth factors that promote their differentiation and survival.(9) Preliminary analysis of the hematopoietic elements within the bone marrow during GCV treatment suggests that both B-cell and erythroid lineages' development is compromised. Interestingly, the earliest hematopoietic progenitors, assessed by phenotypic and functional ability to generate myeloid colonies in methylcellulose cultures, appear to be spared.

These findings suggest that the inhibitory effect of GCV is not caused by the transmission of phosphorylated GCV directly through a gap junction between the osteoblast and adjacent early progenitor(39, 40) but instead results from osteoblast-derived cytokine necessary for progression of the erythroid and B-cell lineages.(41) A direct effect due to low expression of the transgene in the marrow support cells is a possible but unlikely explanation because we have been unable to show transgene activity by immunostaining or Northern analysis of freshly harvested marrow and bone marrow stromal cultures derived for Δtk/+ mice initiate normally in the presence of GCV. In addition, extramedullary hematopoiesis is established and is sufficient to maintain normal peripheral blood cell parameters (unpublished data). Extramedullary hematopoiesis is observed also in the Cbfa1 knockout mice, indicating that blood formation is not absolutely dependent on the presence of differentiated osteoblasts.(42)

The development and activity of the osteoclast also is dependent on the osteoblast. The role that M-CSF, OPGL, and OPG have in regulating osteoclast proliferation and differentiation is now being clarified.(7) M-CSF is necessary for early differentiation of hematopoietic precursors into osteoclast progenitors, whereas OPGL acts more as an inducer of terminal maturation into osteoclasts. The osteoclast-differentiating activity of OPGL is inhibited by OPG.(6) These factors plus interleukin 1 (IL-1) and IL-6 are produced by osteoblasts and stromal cell lines2-4, 43, 44) and normal osteoclastogenesis probably depends on a balance of these and other factors in the bone microenvironment.(45) For example, the level of M-CSF mRNA in both skeleton and calvaria-derived cells in the Cbfa1 knockout mice was similar to wild-type embryos,(12) but the level of OPG was decreased and OPGL was undetectable in femurs and tibias in mutant mice.(4) The ratio of OPGL/OPG in cells cultured from calvaria of the Cbfa1 −/− mice suggests that the lack of osteoblast maturation had a more profound effect on osteoclastogenesis via the deficiency of OPGL production rather than the loss of OPG.(4) Our study would suggest that loss of Col2.3Δtk-expressing osteoblastic cells may have a similar effect on the osteoclastogenesis as seen in Cbfa1 knockout mice. The destruction of early osteoblasts reduce the level of OPGL while the surviving osteocytes continue to produce OPG resulting in the low number of osteoclasts observed in the GCV-treated bones. In contrast, the persistence of osteoclasts in the OC-tk mice would suggests that the OPGL/OPG ratio achieved by the surviving cells within the osteoblast lineage favored OPGL such that osteoclasts persisted despite a reduction of the osteoblast population.(16)

The extensive osteogenic response found in Col2.3Δtk mice 4 weeks after the GCV treatment was terminated differs in the temporal pattern of bone repair observed after physical disruption of the tibial marrow space.(46, 47) In this model, the regeneration process is initiated from a blood clot within the injured marrow space; its peak osteogenic response occurs within 7 days, which is remodeled to form an intact marrow tissue in 2 weeks. Although we have not performed a detailed temporal study of the GCV recovery, it is characterized by a generalized osteogenic response throughout every bone that does not arise from a resorbing blood clot, and it is much slower to resolve. What is less clear is the reason for the exceptional osteogenic response in the Col2.3Δtk mice. One possibility is that the stimulus to generate new bone activates the osteoprogenitor pathway, which may accumulate at a stage before Col2.3 promoter expression. Once the GCV is discontinued this augmented population of cells progresses to full osteoblast differentiation, all be it in a disorganized manner. Additional factors responsible for the exaggerated osteogenic response might include reduced osteoclastic activity or lack of an inhibitory effect that intact hematopoietic cells can exert on bone formation.(48)

In summary, destruction of the early differentiating osteoblast has a major effect on the integrity of hematopoiesis within the bone marrow and the number of osteoclasts that are active within the bone matrix, whereas these effects are not observed when postmitotic mature osteoblasts are targeted for GCV destruction.


We thank Dr. Sandra Weller for the Vero cells infected with herpes virus and we are grateful to Dr. B. Kream for critical reading of the article. This work was supported by National Institutes of Health (NIH) grants R01-AR43457, R01-AR44545, and PO1-AR38933. D.V. is a Fogarty International Fellow (TW05309) and I.K. holds a Michael Geisman Fellowship of the Osteogenesis Imperfecta Foundation.