Estrogen Deficiency Exacerbates Type 1 Diabetes–Induced Bone TNF-α Expression and Osteoporosis in Female Mice

Abstract

Estrogen deficiency after menopause is associated with rapid bone loss, osteoporosis, and increased fracture risk. Type 1 diabetes (T1D), characterized by hypoinsulinemia and hyperglycemia, is also associated with bone loss and increased fracture risk. With better treatment options, T1D patients are living longer; therefore, the number of patients having both T1D and estrogen deficiency is increasing. Little is known about the mechanistic impact of T1D in conjunction with estrogen deficiency on bone physiology and density. To investigate this, 11-week-old mice were ovariectomized (OVX), and T1D was induced by multiple low-dose streptozotocin injection. Microcomputed tomographic analysis indicated a marked reduction in trabecular bone volume fraction (BVF) in T1D-OVX mice (~82%) that was far greater than the reductions (~50%) in BVF in either the OVX and T1D groups. Osteoblast markers, number, and activity were significantly decreased in T1D-OVX mice, to a greater extent than either T1D or OVX mice. Correspondingly, marrow adiposity was significantly increased in T1D-OVX mouse bone. Bone expression analyses revealed that tumor necrosis factor (TNF)–α levels were highest in T1D-OVX mice and correlated with bone loss, and osteoblast and osteocyte death. In vitro studies indicate that estrogen deficiency and high glucose enhance TNF-α expression in response to inflammatory signals. Taken together, T1D combined with estrogen deficiency has a major effect on bone inflammation, which contributes to suppressed bone formation and osteoporosis. Understanding the mechanisms/effects of estrogen deficiency in the presence of T1D on bone health is essential for fracture prevention in this patient population.

T1 diabetes in combination with estrogen deficiency intensifies bone TNF expression, which is linked with exacerbated bone loss, cell death, and marrow adiposity, compared with either condition alone.

Osteoporosis is characterized by decreased bone density and impaired bone architecture (1). There are two main forms of osteoporosis, primary and secondary. Primary osteoporosis is caused by the cessation of estrogen production in menopausal women. It is estimated that one in two women age >50 years will experience an osteoporotic fracture in their lifetime, leading to dependence, depression, and decreased health and life expectancy (2). Secondary osteoporosis results from metabolic diseases and disturbances, organ dysfunction, poor nutrition and lifestyle habits, and being a side effect of various medications (3). Choosing effective therapeutic treatments for bone loss requires knowledge of the pathophysiologic mechanism, which can differ between primary and secondary causes. Additionally, with the growth in effective treatments for chronic diseases, patients are now living longer and faced with conditions associated with aging, leading to a combination of primary and secondary osteoporosis.

Menopause is characterized by the loss of estrogen and leads to primary osteoporosis (1). During the initial years after menopause, the rapid decline in estrogen production results in increased bone remodeling, where both bone formation (osteoblast activity) and resorption (osteoclast activity) are increased (4). These processes, however, are no longer balanced, and bone resorption outweighs bone formation leading to net bone loss and osteoporosis (5–7). Estrogen deficiency during menopause also increases production of inflammatory cytokines, which can stimulate osteoclast activity and decrease osteoclast apoptosis, in addition to inhibiting osteoblast differentiation and activity (8–10).

Type 1 diabetes (T1D), a cause of secondary osteoporosis, is an autoimmune disease in which the β cells of the pancreas are destroyed, resulting in a hypoinsulinemic and hyperglycemic environment (11). Although patients have a longer life span because of exogenous insulin therapy, maintaining euglycemia remains difficult, even under therapeutic vigilance (12, 13). As a result, T1D patients continue to suffer from a number of complications associated with chronic hyperglycemia, including osteoporosis (11). Studies in both animal and human models demonstrate that T1D decreases trabecular and cortical bone density and increases fracture risk (14, 15). This reduction in bone health is primarily caused by effects on the selection, maturation, and activity of osteoblasts (16–18), as evidenced by decreased levels of osteoblast markers of maturity and activity, including alkaline phosphatase and osteocalcin (19–27). Bone formation is linked with metabolic control because formation markers are inversely proportional to glycosylated hemoglobin levels (17, 28, 29). In T1D mouse models, bone inflammatory markers and bone marrow adiposity are increased (4, 5, 26, 30). The role of osteoclasts in T1D osteoporosis is variable (likely dependent on model, time point, and severity), with reports showing no change, decrease, or increase in number and activity (27, 31–34).

Although both causes of osteoporosis, estrogen deficiency (as seen in menopause), and T1D have been researched independently, the increasing life span of T1D patients makes it necessary to examine the combination of the conditions. Clinical studies indicate that postmenopausal women with T1D have decreased bone mineral density (BMD) at the femoral neck (35–37); however, the underlying mechanisms accounting for this response are unknown. Consistent with human studies, we found that the combination of T1D and estrogen deficiency caused by ovariectomy (OVX) decreases femur and vertebral bone volume to a greater extent than either T1D or OVX alone. Our studies further implicate a role for tumor necrosis factor (TNF)–α expression, which is elevated in T1D-OVX bone and negatively correlates with bone volume and osteoblast number, and positively correlates with osteoblast and osteocyte death, all of which are significantly altered in T1D-OVX mice. A role for estrogen deficiency and high glucose in enhancing osteoblast TNF-α expression is suggested. Taken together, our data suggest that postmenopausal women with T1D could be at greater risk for inflammation-induced bone loss compared with postmenopausal women without T1D.

Materials and Methods

Mice and experimental design

Female BALB/c mice were purchased from Jackson Laboratories and allowed to acclimate to the animal facilities for 1 week. Mice were randomly assigned to groups and subsequently either ovariectomized or sham operated at 11 weeks of age. To induce T1D, 1 week after surgery (at 12 weeks of age), mice were intraperitoneally injected with 50 mg/kg body weight of streptozotocin [(STZ) Sigma, St. Louis MO] for 5 consecutive days, and blood glucose levels were measured on the fifth day. In some mice (that did not become diabetic at this dose), an extra single dose of STZ at 80 mg/kg body weight was administered on day 6. Mice with blood glucose >300 mg/dL or an glycosylated hemoglobin A1c (HbA1c) >7% were considered diabetic. Mice that lost 15% of initial body weight (four T1D and five T1D-OVX mice during the final week of the study) were given daily injections of 0.5 mU/g of insulin to keep blood glucose levels >300 but <500 mg/dL and prevent further weight loss. Mice were euthanized 4 weeks after the induction of diabetes. Mice were given Teklad 2019 chow (Envigo, Indianapolis, IN) and water BALB/c ad libitum and were maintained on a 12-hour light/dark cycle. Food and water intake were monitored during the experiment. All animal experimentation described in the submitted manuscript was conducted in accord with accepted standards of humane animal care, as outlined in the ethical guidelines. All animal procedures were approved by the Michigan State University Institutional Animal Care and Use Committee and conformed to National Institutes of Health guidelines.

Serum measurements

Blood was collected at the end of the study and was either placed in EDTA tubes to prevent clotting or allowed to clot at room temperature for 5 minutes. Clotted blood was centrifuged at 10,000 relative centrifugal force for 10 minutes. Serum was removed, aliquoted, and snap frozen in liquid nitrogen and stored at −80°C. Serum tartrate-resistant acid phosphatase (TRAP) 5b and osteocalcin were measured using a Mouse TRAP and OC assay kits (SB-TR103; Immunodiagnostic Systems Inc., Fountain Hills, AZ; and BT – 470, Biomedical Technologies Inc., Stoughton, MA, respectively) per the manufacturer’s protocol. Whole blood in EDTA tubes was used to measure HbA1C levels (Crystal Chem, Downers Grove, IL) according to the manufacturer’s protocol.

Microcomputed tomography bone imaging

Femurs and vertebrae were fixed in formalin and transferred to 70% ethanol after 24 hours. Bones were scanned using an eXplore Locus microcomputed tomography system (GE Health Care) with a voxel resolution of 20 µm obtained from 720 views. Beam strength was set at 80 peak kV and 450 µA with a beam angle increment of 0.5. Each scan consisted of bones from all experimental groups and a calibration phantom bone to maintain consistency throughout the scans. A fixed threshold of 792 was used to separate trabecular bone from bone marrow. A region of interest in the distal femur was analyzed and defined as 1% of the total length proximal to the growth plate and extending 2 mm toward the diaphysis excluding the outer cortical bone. Trabecular bone mineral content (BMC), bone volume fraction (BVF), thickness, spacing, and number values were computed by a MicroView software application (GE Health Care, Wauwatosa, WI) for visualization and analysis of volumetric image data. The third lumbar vertebral body was analyzed for trabecular bone as previously described. Cortical measurements were done on the femur diaphysis using a 2- × 2- × 2-mm region of interest centered midway down the length of the bone using a threshold of 1200 to separate cortical bone from marrow.

Femur histomorphometry and dynamic measures

Femurs were fixed in formalin and transferred to 70% ethanol after 24 hours. Fixed samples were processed on an automated Electron Excelsior tissue processor (Thermo Fisher Scientific) for dehydration, clearing, and infiltration using a routine overnight processing schedule. Samples were embedded in Surgipath-embedding paraffin on a Tissue Tek II-embedding center (Sakura, Torrance, CA). Paraffin blocks were sectioned at 5 µm on a Reichert Jung 2030 rotary microtome (Leica, Wetzlar, Germany).

Cell death, osteoblast, and osteocyte number were determined using a TACS XL Basic In Situ Apoptosis Detection Kit [terminal deoxynucleotidyltransferase-mediated dUTP nick end labeling (TUNEL); Trevigen Inc., Gaithersburg, MD] on femur sections. At least five images at ×40 magnification were taken at the distal metaphysis, excluding the growth plate per section for osteoblast analysis. Cuboidal osteoblasts (aligned in clusters atop osteoid) along the bone surface were counted. The total number of osteoblasts, both TUNEL negative and TUNEL positive (as determined by brown staining) were counted per millimeter of bone surface and expressed as percent TUNEL positive osteoblasts over total osteoblasts. In addition, to assess the number of osteoblasts in the distal femur trabecular area, we took the number of osteoblasts per millimeter of bone surface and multiplied this value by the total bone surface (mm²) determined by microcomputed tomography; this provided a measure of the total number of osteoblasts in the region. For osteocyte quantitation, at least 10 photos at ×20 magnification were taken of the cortical bone in the metaphyseal region of the femur. Total osteocytes, both TUNEL positive and negative, were counted within the cortical bone and expressed relative to the cortical bone area examined. Positive controls included slides which were incubated with nuclease. Counts were done with students blinded to the conditions.

For osteoclast and adipocyte quantification, slides were stained for TRAP activity and counterstained with hematoxylin per manufacturer protocol (387A-IKT; Sigma). At least five images at ×40 magnification were taken at the distal metaphysis per section for osteoclast analysis and at least 10 images at ×20 magnification were taken of the bone marrow for adipocyte measurements. TRAP-positive osteoclast surface along the bone surface was measured and expressed relative to the total bone surface. In addition, the ratio of osteoclasts to osteoblasts (OC/OB) was determined by dividing the number of osteoclasts by the number of osteoblasts. The area of adipocytes in the femur metaphyseal marrow was determined using ImageJ (National Institute of Health, Bethesda, MD). Only adipocytes >30 µm were included in the analyses, as reported previously (26, 30). The adipocyte area was expressed relative to the total regional bone marrow area (excluding trabeculae) and expressed as a percentage of the total area analyzed. Quantitation was performed blinded.

TNF-α protein was examined in bone marrow throughout full-length femur bone sections by immunohistochemistry using a 1:100 dilution of polyclonal anti–TNF-α (ab6671; research resource identifier: AB 305641; Abcam, Cambridge, MA). Femur sections were subsequently counterstained with methylene green and five to 10 digital images were taken at ×10 magnification along the length of the femur. ImageJ software was used to quantify the area of stain (brown) relative to the total area measured. Measures were made blinded.

For dynamic histomorphometric measures of bone formation, mice were injected intraperitoneally with 200 µL of 10 mg/mL calcein (Sigma) dissolved in sterile saline at 7 and 2 days prior to harvest. Vertebrae and humeri were fixed in formalin at the time of harvest and then transferred to 70% ethanol 24 hours later and processed for embedding and sectioning as previously described (6). Vertebrae and humeri were sectioned and examined under fluorescent light. At least five images at ×20 magnification were taken, and the distance between the calcein lines [mineral apposition rate (MAR)] and their lengths along the bone surface [used to calculate bone formation rate (BFR)] were measured blinded to sample conditions.

RNA isolation and analysis

Immediately after euthanasia, tibias were dissected and cleaned of muscle and connective tissue and subsequently snap frozen in liquid nitrogen. Frozen tibias were crushed under liquid nitrogen conditions with a Bessman Tissue Pulverizer (Spectrum Laboratories, Rancho Dominguez, CA) and homogenized in TriReagent (Molecular Research Center, Cincinnati, OH). For bone marrow RNA isolation, the end was removed from freshly isolated femurs, then femurs were spun in a microfuge tube at 5000 relative centrifugal force for 20 seconds; the pelleted marrow was then resuspended in 1 mL of TriReagent. RNA from whole tibia and bone marrow was isolated, and RNA integrity was assessed by formaldehyde-agarose gel electrophoresis. Complementary DNA was synthesized by reverse transcription using Superscript II Reverse transcription Kit and oligo dT (13–19) primers (Invitrogen, Carlsbad, CA) and amplified by real-time polymerase chain reaction with iQ SYBR Green Supermix (BioRad, Hercules, CA). Gene-specific primers were synthesized by Integrated DNA Technologies (Coralville, IA). Hypoxanthine guanine phosphoribosyl transferase messenger RNA levels do not fluctuate between conditions and were used as an internal control. Amplicon specificity was confirmed by melting curve, size, and sequence analysis. Primers for real-time polymerase chain reaction and cycle parameters are listed in Table 1.

Table 1.

Primers

	Primer
HPRT forward	AAG CCT AAG ATG AGC GCA AG
HPRT reverse	TTA CTA GGC AGA TGG CCA CA
RUNX2 forward	GAC AGA AGC TTG ATG ACT CTA AAC C
RUNX2 reverse	TCT GTA ATC TGA CTC TGT CCT TGT G
Osterix forward	CTG CGG AAA GGA GGC ACA AAG AAG
Osterix reverse	GGG TTA AGG GGA GCA AAG TCA GAT
Osteocalcin forward	ACG GTA TCA CTA TTT AGG ACC TGT G
Osteocalcin reverse	ACT TTA TTT TGG AGC TGC TGT GAC
RANKL forward	TTT GCA GGA CTC GAC TCT GGA G
RANKL reverse	TCC CTC CTT TCA TCA GGT TAT GAG
OPG forward	GAA GAA GAT CAT CCA AGA CAT TGA C
OPG reverse	TCC ATA AAC TGA GTA GCT TCA GGA G
TRAP forward	AAT GCC TCG ACC TGG GA
TRAP reverse	CGT AGT CCT CCT TGG CTG CT
FABP4 forward	GCG TGG AAT TCG ATG AAA TCA
FABP4 reverse	CCC GCC ATC TAG GGT TAT GA
TNF-α forward	AGG CTG CCC CGA CTA CGT
TNF-α reverse	GAC TTT CTC CTG GTA TGA GAT AGC AA
IL-10 forward	GGT TGC CAA GCC TTA TCG GA
IL-10 reverse	ACC TGC TCC ACT GCC TTG CT
BAX forward	GAC AGG GGC CTT TTT GCT A
BAX reverse	TGT CCA CGT CAG CAA TCA TC
Bcl-2 forward	GAC AGA AGA TCA TGC CGT CC
Bcl-2 reverse	GGT ACC AAT GGC ACT TCA AG

Abbreviations: FABP4, Fatty acid binding protein 4; HPRT, Hypoxanthine guanine phosphoribosyl transferase; IL, interleukin; OPG, osteoprotegrin.

In vitro osteoblast experiments

Preosteoblast clone 4 MC3T3-E1 cells (CRL-2593; ATCC, Manassas, VA) (passages between 18 and 23) were plated at a density of 19,000 cells in 24 plate wells with 1 mL of α–minimal essential media (α-MEM) with 10% fetal bovine serum (FBS) (Invitrogen and Atlanta Biologicals, Atlanta, GA). Cells were fed every other day with 10% FBS α-MEM. On day 5, media were changed to include α-MEM that is phenol free (Invitrogen) with estrogen-free FBS (Atlanta Biologicals). A portion of cells were treated with estrogen (Sigma) at a final concentration of 1 × 10⁻⁸ M. Cells were subsequently treated 24 hours later by addition of glucose (30 mM final concentration) and/or TNF-α (10 ng/mL; R & D Systems, Minneapolis, MO). Cells were harvested 24 hours later with TriReagent (Molecular Research Center Inc.), and RNA was extracted and analyzed as previously described.

Statistical analyses

All measurements are presented as mean ± standard error of the mean. Significant outliers were removed using the Rout test (in some cases one outlier was found in a group). One- and two-way analysis of variance (ANOVA) and Pearson correlation analyses were performed using GraphPad Prism software version 7 (GraphPad, San Diego, CA); One-way ANOVAs were followed by Fisher least significant difference test. P ≤ 0.05 was considered significant, and values ≤0.01 were considered highly significant.

Results

To determine if T1D in the presence of estrogen deficiency has a greater impact on bone health than either condition alone, 11-week-old female mice were either sham-operated (sham) or ovariectomized and then treated with vehicle (control) or STZ (which induces T1D). General body parameters were measured to assess effects on body and fat weight, metabolic dysregulation, and uterine size. As expected, T1D mice, both sham and OVX, had significantly elevated blood glucose and HbA1c levels, indicating metabolic dysregulation (Table 2). Interestingly, OVX mice also had increased HbA1c levels, though they were <7% (>7% would be considered diabetic); however, T1D in the presence of estrogen deficiency (OVX) mice did not further raise HbA1c levels. Examination of body and organ weights demonstrated that all diabetic mouse conditions (±OVX) displayed significantly decreased body mass and fat pad (inguinal and retroperitoneal) mass compared with control mice (Table 2). In contrast, estrogen deficiency increased body mass and retroperitoneal fat pad mass in sham mice, but not in T1D mice. As expected, all OVX mice displayed markedly reduced uterine mass confirming estrogen deficiency (Table 2), whereas the average uterine size in sham mice was slightly lower in T1D mice, this was not important and could be accounted for by the stage of estrus cycle at the time of harvest (data not shown).

Table 2.

General Mouse Parameters

Parameter	Control		Diabetic
	Sham (n = 10)	OVX (n = 12)	Sham (n = 9)	OVX (n = 12)
Body mass (g)	21.5 ± 0.5	23.2 ± 0.4	17.7 ± 0.4a,b	18.2 ± 0.6a,b
Blood glucose (mg/dL)	251 ± 24	254 ± 19	483 ± 28a,b	441 ± 20a,b
HbA1c	4.1 ± 0.3	4.7 ± 0.2	8.6 ± 0.6a,b	8.4 ± 0.5a,b
Subcutaneous fat (% body mass)	0.87 ± 0.05	1.02 ± 0.04	0.37 ± 0.03a,b	0.40 ± 0.03a,b
Retroperitoneal fat (% body mass)	0.28 ± 0.02	0.37 ± 0.04	0.16 ± 0.06c	0.11 ± 0.02a,b
Uterus (% body mass)	0.55 ± 0.06	0.10 ± 0.01a	0.43 ± 0.06c	0.10 ± 0.01a,c

Values represent mean ± standard error.

^aP < 0.05 with respect to sham control.

^bP < 0.05 with respect to OVX control.

^cP < 0.05 with respect to T1D-sham.

Microcomputed tomography analyses of the distal femoral trabecular BVF revealed a significant decrease in OVX (52%, P < 0.0001) and T1D (57%, P < 0.0001) as expected, but more importantly the BVF in combined T1D-OVX mice (84%, P < 0.0001) was significantly less compared with mice with either T1D (by 63%, P < 0.0001) or OVX (by 66%, P < 0.0001) [Fig. 1(a) and 1(b)]. Two-way ANOVA identified a highly significant interaction between T1D and OVX (P < 0.0001). Consistent with this finding, BMD, BMC, trabecular thickness, and trabecular number were all significantly decreased in OVX and T1D animals compared with controls, whereas T1D-OVX animals again displayed the greatest reduction, being significantly decreased with respect to sham control, OVX, and T1D mice (Table 3). Trabecular spacing was significantly increased in OVX, T1D, and T1D-OVX compared with the sham control, with T1D-OVX being significantly higher than all conditions (Table 3). Femoral cortical analysis revealed no change in OVX mice, whereas significantly decreased BMD and BMC was observed in T1D mice. T1D-OVX mice had significantly decreased cortical thickness, outer perimeter, marrow area, cortical area, BMD, and BMC compared with controls (Table 3).

Figure 1.

Estrogen deficiency exacerbates type 1 diabetes–induced trabecular bone loss. Eleven-week-old females underwent, sham or OVX operations. At 12 weeks of age, T1D (diabetes) was induced by streptozotocin injection, whereas controls were injected with saline. At 16 weeks of age, bones were imaged by microcomputed tomography. (a) Distal femur trabecular BVF. (c) L3-vertebral trabecular BVF. (b and d) Representative isosurface images of distal femoral and vertebral trabecular bone, respectively. Values represent mean ± standard error (n = 10 for intact control group; n = 9 for intact T1D group; n = 12 for both OVX control and T1D group). *P < 0.05 with respect to intact control; #P < 0.05 with respect to OVX control; ^P < 0.05 with respect to T1D intact.

Table 3.

Femoral and Vertebral Bone Parameters

Parameter	Control		Diabetic
	Sham (n = 10)	OVX (n = 12)	Sham (n = 9)	OVX (n = 12)
Femur trabecular
BV/TV	30.1 ± 1.7	13.4 ± 1.1a	10.5 ± 1.5a	4.4 ± 0.4a,b,c
BMD (mg/mL)	297.0 ± 6.4	209.0 ± 8.1a	186.8 ± 11.2a	136.7 ± 6.8a,b,c
BMC (mg)	0.43 ± 0.02	0.35 ± 0.01a	0.32 ± 0.01a	0.25 ± 0.02a,b,c
Tb. Th. (µm)	56.8 ± 1.5	42.1 ± 1.0a	38.0 ± 2.0a	31.6 ± 1.0a,b,c
Tb. N. (1/mm)	5.26 ± 0.21	3.17 ± 0.19a	2.68 ± 0.24a	1.38 ± 0.11a,b,c
Tb. Sp. (µm)	135.8 ± 7.5	286.0 ± 19.9	358.7 ± 36.0a	754.0 ± 74.2a,b,c
Vertebral trabecular
BV/TV	40.2 ± 2.4	40.4 ± 2.5	28.8 ± 2.7a,b	21.6 ± 1.4a,b
BMD (mg/mL)	286.6 ± 9.8	284.9 ± 10.41	253.1 ± 13.3a	200.6 ± 6.3a,b,c
BMC (mg)	0.46 ± 0.03	0.45 ± 0.04	0.38 ± 0.03	0.37 ± 0.03
Tb. Th. (µm)	58.9 ± 2.6	56.8 ± 2.8	44.5 ± 3.0a	37.4 ± 1.3a,b
Tb. N. (1/mm)	6.79 ± 0.26	7.11 ± 0.19	6.40 ± 0.25	5.72 ± 0.23a,b
Tb. Sp. (µm)	89.8 ± 6.0	84.9 ± 5.2	113.6 ± 8.5	141.3 ± 9.5a,b
Femur cortical
Mean thickness (mm)	0.255 ± 0.004	0.261 ± 0.005	0.240 ± 0.006	0.229 ± 0.004a,b
Inner perimeter (mm)	2.76 ± 0.03	2.76 ± 0.02	2.84 ± 0.03	2.84 ± 0.03
Outer perimeter (mm)	4.37 ± 0.03	4.40 ± 0.03	4.46 ± 0.11	4.29 ± 0.03
Marrow area (mm2)	0.54 ± 0.01	0.54 ± 0.01	0.57 ± 0.01	0.58 ± 0.02
Cortical area (mm2)	0.87 ± 0.02	0.90 ± 0.02	0.82 ± 0.02b	0.78 ± 0.02a,b
BMD (mg/mL)	1039 ± 22	1054 ± 12	983 ± 15b	985 ± 18a,b
BMC (µg)	18.2 ± 0.4	18.9 ± 0.4	16.4 ± 0.5a,b	15.5 ± 0.5a,b

Values represent mean ± standard error (n ≥ 5 per group).

Abbreviations: BV/TV, bone volume/total volume; Tb. N., trabecular number; Tb. Sp., trabecular spacing; Tb. Th., trabecular thickness.

^aP < 0.05 with respect to sham control.

^bP < 0.05 with respect to OVX control.

^cP < 0.05 with respect to T1D-sham.

Lumbar vertebrae (L3) were also examined by microcomputed tomography to establish if the observed femur trabecular bone changes occur at another bone site. Although OVX did not alter vertebral trabecular BVF compared with sham control mice, a significant decrease was observed in T1D mice compared with controls (24%, P < 0.05), and between T1D-OVX mice compared with control (51%, P < 0.0001), OVX (47%, P < 0.0001), and T1D (35%, P < 0.01) mice [Fig. 1(c) and 1(d)]. Two-way ANOVA, however, did not indicate an interaction between T1D and OVX at this site, despite the marked reduction in BVF in the T1D-OVX mice. In addition, T1D-OVX mice displayed significantly decreased vertebral BVF, BMD, trabecular thickness, and trabecular number, and increased trabecular spacing (Table 3).

Previous work has shown that T1D decreases osteoblast lineage selection, maturation, and viability (16, 17, 38–41). Therefore, we examined the impact of T1D in combination with OVX on several osteoblast RNA and serum markers. RNA expression analyses revealed a significant decrease in RUNX2 (early osteoblast differentiation gene) in the T1D mice with respect to OVX controls and T1D-OVX animals compared with the sham and OVX controls [Fig. 2(a)]. Osterix (osteoblast maturation marker) messenger RNA was significantly decreased in OVX (P < 0.01) and T1D-sham (P < 0.01) and was further decreased in T1D-OVX (P < 0.001) animals with respect to control sham-operated animals (two-way ANOVA, P = 0.1592) [Fig. 2(b)]. Osteocalcin (osteoblast differentiation marker) messenger RNA and serum levels were significantly decreased in both T1D sham (P < 0.05) and T1D-OVX (P < 0.05) conditions [Fig. 2(c) and 2(d)]. Neither measure was affected by OVX alone. Histomorphometric quantitation of osteoblast number per millimeter bone surface indicated a significant reduction in T1D-OVX mice [Fig. 2(e)], whereas T1D and OVX trended to reduce osteoblast number (P = 0.07). Calculation of the total number of osteoblasts on the total femur trabecular surface indicated significantly less osteoblasts per millimeter of bone in OVX (P < 0.0001), T1D (P < 0.05), and T1D-OVX (P < 0.0001) conditions with respect to sham controls. Additionally, the T1D-OVX had significantly lower total number of osteoblasts than either OVX (P < 0.0001) or T1D (P < 0.0001) animals [Fig. 2(f)], which parallels BVF data shown in Fig. 1.

Figure 2.

Type 1 diabetic, estrogen-deficient mice display lowest osteoblast marker levels. (a–c) Gene expression of osteoblast bone markers measured in RNA extracted from whole tibia (RUNX2, Osterix, and osteocalcin). (d) Osteocalcin protein levels in serum. (e) Osteoblast number per millimeter bone surface as determined histomorphometrically. (f) Calculated total osteoblast number per trabecular bone surface of the distal femur metaphyseal region. (g) Representative images (×40 magnification) of osteoblasts. Arrows indicate examples of osteoblasts that were quantified. Values represent mean ± standard error (n = 10 for intact control group; n = 9 for intact T1D group; n = 12 for both OVX control and T1D group). *P < 0.05 with respect to intact control; #P < 0.05 with respect to OVX control; ^P < 0.05 with respect to T1D intact. HPRT, hypoxanthine guanine phosphoribosyl transferase.

Dynamic histomorphometry measures were used to detect if suppressed osteoblast activity contributes to the identified bone loss in vertebrae (L3-4) and long bone (humerus). In correlation with the vertebral bone density analyses, the OVX, T1D, and T1D-OVX mice had significantly decreased MAR with respect to control animals (P < 0.01) [Fig. 3(a)]. Both T1D groups (sham and OVX) had significantly lower MAR in comparison with OVX animals [Fig. 3(a)] (P < 0.01). Calculation of BFR indicated a decrease in both the T1D and T1D-OVX groups with respect to the sham control [Fig. 3(b)] (P < 0.05), but no change was observed in the OVX groups. Long bone dynamic measures in the humerus showed a significant decrease in both T1D and T1D-OVX MAR with respect to sham animals. Furthermore, T1D-OVX mice had a significantly lower MAR than OVX controls [Fig. 3(c)]. Consistent with long bone BVF measurements, both T1D conditions had significantly lower BFRs than either the sham or OVX controls [Fig. 3(d)]. Additionally, T1D-OVX mice had lower BFRs than T1D alone [Fig. 3(d) and 3(e)]. This indicates that T1D and OVX suppress bone formation in long bones to a greater extent than either condition alone.

Figure 3.

T1D-OVX condition decreases dynamic markers of bone formation in both the vertebrae and humeri. (a and b) Calculated vertebral MAR and BFR. (c and d) Calculated humerus MAR and BFR. (e) Representative fluorescent microscopy images of calcein pulses located near the bone surface. Values represent mean ± standard error (n = 10 for intact control group; n = 9 for intact T1D group; n = 12 for both OVX control and T1D group). *P < 0.05 with respect to intact control; #P < 0.05 with respect to OVX control; ^P < 0.05 with respect to T1D intact.

To examine the effect of T1D-OVX on osteoclast activity, we analyzed osteoclast markers in serum, tibia messenger RNA, and femoral histomorphometry. TRAP messenger RNA levels were not affected by T1D (compared with control), but were significantly increased in T1D-OVX animals with respect to sham (P < 0.01), OVX controls (P < 0.0001), and T1D animals [Fig. 4(a)]. Two-way ANOVA indicated an interaction between T1D and OVX (P < 0.0027). Serum TRAP5b, an indicator of osteoclast activity, was not affected by OVX but was increased in T1D-sham and T1D-OVX animals compared with the sham controls (P < 0.01) and OVX mice (P < 0.01) [Fig. 4(b)]. We further examined regulators of osteoclast activity and found that the ratio of expression levels of receptor activator of nuclear factor κ-B ligand (RANKL) (critical in osteoclast differentiation) to osteoprotegrin (an inhibitor of RANKL) was significantly increased in T1D and T1D-OVX animals with respect to sham controls [Fig. 4(c)]. Histomorphometric analyses did not detect a change in osteoclast surface between conditions [Fig. 4(d)]. However, when expressed as OC/OB relative to the total bone surface area, T1D-OVX mice exhibited significantly higher OC/OB compared with sham control (P < 0.05) and T1D groups (P < 0.01) [Fig. 4(e)].

Figure 4.

T1D-OVX mice display altered osteoclast markers. (a) TRAP expression levels in RNA extracted from whole tibias. (b) TRAP5b levels in mouse serum. (c) RANKL (an activator of osteoclastogenesis) expression relative to osteoprotegrin (OPG) (an inhibitor of osteoclastogenesis) messenger RNA levels. (d) Percentage of osteoclast surface per millimeter bone surface. (e) Calculated OC/OB. (f) Representative images (×40 magnification) of TRAP-stained osteoclasts, with arrows indicating examples of osteoclasts quantified. Values represent mean ± standard error (n = 10 for intact control group; n = 9 for intact T1D group; n = 12 for both OVX control and T1D group). *P < 0.05 with respect to intact control; #P < 0.05 with respect to OVX control; ^P < 0.05 with respect to T1D intact. HPRT, hypoxanthine guanine phosphoribosyl transferase.

Osteoblasts and adipocytes are both derived from mesenchymal stem cells. Selection of adipocytes over osteoblasts will decrease the number of osteoblasts available to remodel bone. Both OVX and T1D have been previously shown to increase bone marrow adiposity (30, 42); thus, we examined bone marrow adipocyte numbers in our study. Histomorphometry analyses indicated an increase in the percentage of adipocyte/total marrow area in both OVX and T1D mice compared with controls, but an even greater increase occurred in T1D-OVX animals compared with sham controls (P < 0.001) [Fig. 5(a) and 5(b)]. Marrow Fatty acid binding protein 4 expression displayed a similar pattern to marrow adiposity [Fig. 5(c)]. Consistent with the reciprocal relationship typically seen between osteoblasts and adipocytes, the observed increase in adipocytes paralleled the decrease in bone volume and osteoblast markers and numbers (Figs. 1 and 2).

Figure 5.

Bone marrow adiposity is increased in estrogen deficient diabetic mice. (a) Percentage of adipocyte area to total bone marrow area. (b) Representative marrow images (×40 magnification) displaying adipocyte differences. White circles are adipocytes. (c) Fatty acid binding protein 4 (FABP4) expression levels measured in RNA extracted from femur bone marrow. Values represent mean ± standard error (n = 10 for intact control group; n = 9 for intact T1D group; n = 12 for both OVX control and T1D group). *P < 0.05 with respect to intact control; #P < 0.05 with respect to OVX control; ^P < 0.05 with respect to T1D intact. HPRT, hypoxanthine guanine phosphoribosyl transferase.

Previous studies indicate that T1D and OVX promote inflammation in bone, specifically by elevating TNF-α, which contributes to bone loss (4, 5, 7–10). Therefore, we hypothesized that the combination of T1D and OVX could greatly enhance TNF-α levels, which results in greater bone loss than either condition alone. To test this, we measured TNF-α expression levels in whole bone RNA extracts. Figure 6(a) indicates that T1D-OVX leads to a significantly greater increase in bone TNF-α expression than T1D or any other condition. The ratio of TNF-α/IL-10 expression, a marker of inflammation, supports greater inflammation in T1D-OVX bone [Fig. 6(b)], consistent with this condition having the greatest bone loss. Using Pearson correlation analyses to determine if there is a link between TNF-α and bone loss, we identified a highly significant (P = 0.0003) negative correlation between femur BVF and bone TNF-α levels [Fig. 6(c)]. A significant negative correlation was also observed between bone TNF-α levels and osteoblast number [Fig. 6(d)]. The increase in TNF-α expression was further confirmed by immunohistochemical analyses, which demonstrated a significant increase in TNF-α protein levels in T1D-OVX femur marrow compared with sham, OVX, or T1D alone [Fig. 6(e)]. To determine if TNF-α expression is elevated in osteoblasts to a greater extent under estrogen deficient and high glucose conditions, MC3T3 osteoblasts were exposed to estrogen deficiency, high glucose (30 mM), and inflammation (TNF-α treatment) separately or together, and then TNF-α expression was measured. Although glucose treatment alone did not affect TNF-α expression compared with controls (data not shown), TNF-α treatment in the presence of elevated glucose and estrogen deficiency caused a significant increase in TNF-α expression that was greater than in cells cultured with estrogen [Fig. 6(f)]. This further supports the combinatorial effects of estrogen deficiency and high glucose on bone TNF-α levels, which promote corresponding bone loss.

Figure 6.

Bone TNF-α expression is increased by T1D-OVX conditions in vivo and in vitro and correlates to BVF and osteoblast numbers. (a) TNF-α RNA levels measured from whole tibia. (b) Ratio of the messenger RNA levels of TNF-α to interleukin (IL)-10. (c and d) Pearson correlation analyses of TNF-α RNA levels relative to (c) femoral BVF and (d) osteoblast number per square millimeter bone surface. Values represent mean ± standard error (n = 10 for intact control group; n = 9 for intact T1D group; n = 12 for both OVX control and T1D group). (e) Quantification of TNF-α protein in the femoral bone marrow. Below the graph are representative images (×40 magnification) of quantified immunohistochemical stain in the bone marrow. Values represent mean ± standard error (n = 8 for intact control group; n = 7 for intact T1D group; n = 6 for both OVX control and T1D group). *P < 0.05 with respect to intact control; #P < 0.05 with respect to OVX control; ^P < 0.05 with respect to T1D intact. (f) After 5 days in culture, MC3T3-E1 osteoblasts were grown with or without estrogen. After 24 hours, cells were treated with (c) vehicle, TNF-α (10 ng/mL) or glucose (30 mM) + TNF-α (10 ng/mL). RNA was extracted 24 hours later for TNF-α RNA expression analyses. Values represent mean ± standard error (5 separate experiments per group). *P < 0.05. HPRT, hypoxanthine guanine phosphoribosyl transferase; OB, osteoblast.

Bone inflammation can lead to increased osteoblast and osteocyte death, which can contribute to bone loss. Examination of cell death markers in bone indicated the highest proapoptotic BAX/Bcl-2 ratio in the T1D-OVX mouse bones [Fig. 7(a)]. Therefore, we examined cell death. In addition to a reduction in osteoblast number [Fig. 2(e)], we identified fewer osteocytes in T1D-OVX bone compared with any other condition [Fig. 7(b)]. TUNEL staining of bone sections further indicated more dying osteoblasts and osteocytes in the T1D-OVX mice [Fig. 7(d) and 7(f)]. The percentage of TUNEL-positive osteocytes and osteoblasts was strongly correlated with bone TNF-α levels [Fig. 7(e) and 7(g), suggesting a link between T1D-OVX, increased TNF-α, decreased osteoblast and osteocyte viability, decreased bone formation, and ultimately bone loss.

Figure 7.

Osteoblast and osteocyte death is increased in estrogen-deficient, T1D mice. (a) Ratio of RNA expression of proapoptotic marker, BAX, to the antiapoptotic marker, Bcl-2, from whole tibia. (b) Number of osteocytes per square millimeter of cortical bone in the distal femur metaphyseal region. (c) Representative images of TUNEL-stained osteocytes (×20 magnification) with arrows indicating examples of quantified osteocytes. (d) Percentage of TUNEL-stained osteocytes. (f) Percentage of TUNEL-stained osteoblasts. (e and g) Pearson correlation analyses of TUNEL-stained osteocytes and osteoblasts with TNF-α staining. Values represent the mean ± standard error (n = 10 for intact control group; n = 9 for intact T1D group; n = 12 for both OVX control and T1D group). *P < 0.05 with respect to intact control; #P < 0.05 with respect to OVX control; ^P < 0.05 with respect to T1D intact.

Discussion

Both T1D and estrogen deficiency lead to osteoporosis; however, few studies have examined the mechanisms of how the combination of T1D under estrogen-deficient conditions impacts bone remodeling and physiology. Our findings indicate that the interaction of T1D with OVX leads to a greater loss of trabecular bone volume in both femoral and vertebral bone. T1D-OVX mice exhibited worse trabecular parameters, and in some cases cortical bone parameters, compared with sham, T1D, or OVX conditions alone. BVF strongly correlated (negatively) with bone TNF-α expression, which was greatest in T1D-OVX mice, implicating inflammation as a key factor accounting for the greatest bone loss in T1D-OVX mice. Bone TNF-α also correlated with osteoblast and osteocyte death, suggesting that the elevated inflammation in T1D-OVX mice contributes to bone loss through the suppression bone formation, consistent with the observed suppression in MAR and BFR.

Clinically, in female T1D patients, bone loss and increased fracture risk are seen in both pre- and postmenopausal women and in trabecular and cortical bone regions (11–13). Previous studies indicate that postmenopausal T1D women have decreased BMD at the femoral neck (35–37) and a 6.9- to 12-fold greater risk of hip fracture compared with nondiabetic women (14, 15). Another study by Hothersall et al. (14) demonstrated that absolute hip fracture risk in T1D is greatest at older ages and is greater for T1D women in all age groups >50 years compared with T1D men. These findings are consistent with our studies that demonstrate substantial bone loss in estrogen-deficient T1D mice. Other clinical studies comparing skeletal health between postmenopausal T1D and postmenopausal type 2 diabetes (T2D) patients found vertebral and femoral BMD was increased in postmenopausal women with T2D in comparison with those with T1D (36, 43), indicating that bone loss during estrogen deficiency is greater in T1D compared with T2D patients.

Consistent with our finding that T1D-OVX mice lose the greatest amount of bone, previous studies using high-resolution dual-energy X-ray absorptiometry analyses of T1D, OVX, or T1D-OVX rats indicated that femoral BMD displayed the greatest reduction in the T1D-OVX group (44–46). Interestingly, dual-energy X-ray absorptiometry analysis indicated that rat vertebral BMD (and BMC) in the T1D-OVX rats was not significantly different than T1D alone; this finding is in contrast with our study where we do see a significant decrease in vertebral trabecular BVF and BMD compared with OVX and T1D groups. Differences between studies could result from different imaging approaches and differences between rat and mouse responses in the time frame studied. It should be noted that we carried out our studies in BALB/c mice because of their higher bone density from which to observe a decrease in response to estrogen deficiency and T1D. Variations between our studies and other studies could result from mouse strain differences as well. Previously, we have shown that male T1D BALB/c mice lose bone and exhibit marrow adiposity similar to C57BL/6 mice (17); however, we have not assessed strain differences in female mice. In addition to trabecular analyses, we also examined cortical bone measures and demonstrated that T1D-OVX decreases cortical bone mean thickness and area. Although not directly looking at cortical bone, another study examined the effect of T1D-OVX on bone matrix and found a marked reduction in heparin sulfate and chondroitin sulfate content (46). This study included treatment with insulin or estradiol which prevented the matrix changes (46) and indicated the critical role of estrogen and normal glucose metabolism in maintaining bone matrix and density. Our study agrees with estrogen and glucose metabolism playing a role in maintaining healthy bone.

T1D decreases osteoblast lineage selection and differentiation from mesenchymal stem cells and decreasing osteoblast activity, resulting in reduced bone formation and subsequently bone loss (8, 9, 17, 25). Our studies further support that T1D has a major impact on bone anabolic properties. Specifically, we found that T1D suppresses osteoblast RNA and functional markers and increases marrow adiposity, whereas estrogen deficiency affected only some of these parameters. Although there was a trend for T1D-OVX mice to display the lowest levels of osteoblast markers, only T1D-OVX osteoblast number and humerus BFR were significantly different from T1D sham and control OVX mice. These findings are consistent with studies in rats where estrogen deficiency was associated with increased bone turnover in both control and diabetic rats and bone formation was significantly reduced in diseased vs control ovariectomized rats (18). Given the reciprocal relationship between bone formation and adiposity (19), we also examined marrow adiposity. Both OVX and T1D mice displayed increased marrow adiposity, consistent with previous reports (30, 47–49). In addition, we identified that T1D-OVX mice display the greatest increase in marrow adiposity compared with sham, OVX, or T1D animals. This correlated with the greatest bone loss in T1D-OVX.

Estrogen deficiency–induced bone loss is predominantly a result of increased osteoclast activity and life span, which typically occurs during the first few years after menopause (50, 51). In the current study, markers of osteoclast activity (TRAP and RANKL), or the OC/OB in OVX mice, were not increased. Changes in osteoclast activity typically occur early after ovary removal (20); therefore, it is highly possible that changes occurred at an earlier time point than presented in this study. In contrast, the T1D-OVX group had increased expression of TRAP, in both serum and tibia messenger RNA expression, increased RANKL/osteoprotegrin ratio, and an increase in OC/OB. In contrast with our studies, T1D-OVX rats displayed a decrease in serum TRAP with respect to either T1D or OVX alone (46); this could be because of a variety of differences, including the animal model, bone remodeling rates, and animal stage. Interestingly, when expressed as OC/OB, the diabetic sham mice do display a lower resorption/formation ratio compared with OVX mice, whereas the T1D-OVX mice displayed a greater OC/OB than T1D alone or controls. Overall, neither T1D or OVX affected osteoclast markers consistently.

Inflammation, marked by increased serum proinflammatory cytokine levels, is linked to bone loss (21–28). Proinflammatory cytokines can suppress osteoblast activity while at the same time increase osteoclast activities. Menopause is known to be associated with inflammation, indicating the key role that sex steroids play in inflammatory regulation (29, 30). Correspondingly, in the mouse model, OVX has been shown to increase in bone proinflammatory markers and loss (31–36), consistent with our findings of elevated TNF-α. Similarly, T1D is an inflammatory disease, where inflammation is first targeted at insulin-secreting pancreatic β cells, but then inflammation is maintained likely as a result of metabolic dysregulation (37, 38). Proinflammatory cytokines, such as TNF-α, have also been found to be elevated in T1D mouse models in serum and bone (4, 5, 10), consistent with our findings.

Although inflammation is studied in individual conditions, few studies examine the impact of multiple inflammatory states at the same time. With better therapeutics, women are living longer and experiencing menopause in addition to other preexisting diseases that they may have. Our studies show that the impact of T1D in the presence of estrogen deficiency on bone inflammation is greater than either condition alone. This is consistent with studies indicating that postmenopausal women with T1D have increased serum interleukin-6 levels (38). In our studies, we have found that T1D-OVX mice display the highest levels of bone TNF-α, which negatively correlated with the level of bone loss and osteoblast and osteocyte death. TNF-α has been demonstrated to be particularly pathogenic to bone, causing a decrease in osteoblast viability and bone density and increased adipogenesis (39–41), similar to our findings in the T1D-OVX group. Our cell culture studies further demonstrate that osteoblast TNF-α expression can be exacerbated under the combined conditions of inflammation, high glucose, and estrogen deficiency. Thus, our data support a role for estrogen deficiency maximizing T1D bone TNF-α expression, which exacerbates T1D bone loss. Given that past studies suppressing TNF-α signaling and/or inflammation can enhance osteoblast/bone health (4, 20, 26, 42), it is likely that anti-inflammatory agents will benefit the severe T1D-OVX bone pathology. Further understanding of how T1D and OVX work to impact bone will lead to better prevention techniques and new therapeutic targets to stop bone loss and fractures in this patient population.