Short-term exposure to low-carbohydrate, high-fat diets induces low bone mineral density and reduces bone formation in rats

Abstract

Low-carbohydrate, high-fat (LC-HF) diets are popular for inducing weight loss in adults and are also used as part of a treatment for children with epilepsy. However, potential risks and side effects remain controversial. We investigated effects of LC-HF diets on growth, bone mineral density (BMD), and turnover in growing rats fed for 4 weeks either normal chow (CH, 9% fat, 33% protein, and 58% carbohydrates), LC-HF-1 (66% fat, 33% protein, and 1% carbohydrates), or LC-HF-2 (94.5% fat, 4.2% protein, and 1.3% carbohydrates). Rats fed LC-HF diets accumulated significantly more visceral and bone marrow fat and showed increased leptin but decreased insulin-like growth-factor 1 (IGF-1). Both LC-HF diets significantly decreased body length (nose to rump), but lengths of humerus, tibia, and femur were significantly reduced with LC-HF-2 only. Peripheral quantitative computed tomography (pQCT) and micro-CT (µCT) independently revealed significant reductions in BMD of tibiae in both LC-HF groups, and tibial maximum load was impaired. Bone-formation marker N-terminal propeptide of type I procollagen was reduced in sera of LC-HF groups, whereas bone resorption marker CrossLaps remained unchanged. Real-time PCR analysis revealed significant reductions by 70% to 80% of transcription factors influencing osteoblastogenesis (Runx2, osterix, and C/EBPβ) in bone marrow of rats fed LC-HF diets. In conclusion, both LC-HF diets impaired longitudinal growth, BMD, and mechanical properties, possibly mediated by reductions in circulating IGF-1. Serum bone-formation markers as well as expression of transcription factors influencing osteoblastogenesis were reduced. This might indicate a lower rate of mesenchymal stem cells differentiating into osteoblasts, thus explaining reduced bone formation with LC-HF diets. © 2010 American Society for Bone and Mineral Research

Introduction

Low-carbohydrate, high-fat (LC-HF) diets have become popular in recent years and are thought to be an efficient alternative to low-fat diets for induction of weight loss in overweight subjects.1 Despite frequent use, potential side effects of these diets remain unclear or controversial. A metabolic study in healthy subjects by Reedy and colleagues suggested that LC, high-protein diets could generate an increased risk for bone loss,2 whereas another investigation analyzing the effect of LC-HF diets in adults did not find a significant impact of these diets on bone turnover.3 Ketogenic LC-HF diets are also used therapeutically to alleviate epileptic seizures in children.4 Recently, it has been shown that epileptic children treated with ketogenic diets show decreased height Z-scores as soon as 6 months after the onset of the dietary intervention. Furthermore, a progressive loss of bone mineral content has been reported in children treated with these diets.6 It is well known that nutrition can influence growth and bone metabolism substantially. Standard HF diets with normal carbohydrate content induced adverse effects on bone morphology and mechanics in rats.7, 8 Furthermore, caloric restriction and low dietary protein intake have been shown to negatively influences cortical bone mass, femoral bone mineral density, and fracture load.9, 10 However, mechanistic data on the specific effects of LC-HF diets on bone metabolism are missing. In a previous study with 12-week-old rats fed with an extremely high-fat LC-HF diet,11 we reported two major findings. First, rats fed with the LC-HF diet gained less weight but accumulated significantly more visceral fat compared with controls. Second, feeding of this diet resulted in a decrease in circulating growth hormone (GH) and insulin-like growth factor 1 (IGF-1).

In concert with sex steroids and parathyroid hormone,12, 13 the GH/IGF system potently stimulates bone growth by activating the osteoblast differentiation program.14 IGF-1-deficient mice exhibit serious skeletal malformations, delayed mineralization, reduced chondrocyte proliferation, and increased chondrocyte apoptosis.15 Also in adult animals, IGF-1 greatly influences bone growth and metabolism. In rats, IGF-1 treatment prevented loss of bone volume during skeletal unloading16 and was effective in inducing gene expression for osteoblastic markers in young and old rats. GH itself also affects bone growth. Local injections of GH at the site of the epiphysial growth plate stimulated unilateral bone growth,18 and GH treatment spurred longitudinal growth and bone formation.19, 20 In addition to GH and IGF-1, the six IGF-binding proteins also have various, in part IGF-independent, effects on bone growth.21, 22

In vertebrates, bones are constantly renewed by the physiologic process called bone remodeling,23 a balanced system of bone formation and bone resorption. The rate of bone formation largely depends on the number and activity of osteoblasts. Interestingly, bone marrow osteoblasts and adipocytes share a common precursor cell lineage. Therefore, increased adipocytogenesis in bone marrow can result in less mesenchymal cells differentiating into osteoblasts and thus impair bone formation.24 In recent years, several transcription factors, among others, Runt-related transcription factor 2 (Runx2) and osterix, have been identified that are capable of influencing the fate of the mesenchymal precursor cells toward either osteoblasts or adipocytes.25, 26

Until now, there are no specific data on effects of LC-HF diets on bone growth and quality, but the changes in the GH/IGF axis make it very likely that there is an impact of these diets. We therefore analyzed bone growth, bone mineral density (BMD), and bone turnover in rats fed with LC-HF diets in the current study. In addition, since our rats showed an obese phenotype, we also were interested in investigating accumulation of fat in bone marrow and a potential influence on expression of transcription factors known to influence the fate of mesenchymal stem cells in bone marrow. To distinguish between effects induced by the near absence of carbohydrates in LC-HF diets and effects mainly caused by the low protein content, we included two different LC-HF diets in this study: LC-HF-1 has a relatively high protein content and thus reflects a macronutrient composition commonly used in “Atkin's style” diets in humans, whereas LC-HF-2 is extremely high in fat and comparably low in protein, a ketogenic-like diet used for treatment of epilepsy in children.

Materials and Methods

Animal husbandry and feed composition

Twenty-four male Wistar rats (Harlan-Laboratories, Borchen, Germany) were housed in individual cages in a thermoneutral environment (21.3 ± 0.59°C, humidity 70 ± 1.2%) maintained on a 12-hour light-dark cycle throughout the study (lights on at 11:00 p.m. and off at 11:00 a.m). All animals received ad libitum access to water and standard laboratory chow (CH) for the first 10 days following delivery to allow acclimation to the new environment. Body weight and 24 hour food intake were measured daily (Sartorius Competence CP2201, Goettingen, Germany) to the nearest 0.1 g one hour before the onset of the dark period. At the end of the acclimation period, rats had reached an age of 4 weeks, were matched for body weight, and were divided into the three dietary groups (n = 8 rats). Laboratory standard chow (CH) consisted of of 9% fat, 33% protein, and 58% carbohydrates; the LC-HF-1 diet consisted of 66% fat, 33% protein, and 1% carbohydrates, and the LC-HF-2 consisted of 94.5% fat, 4.2% protein, and 1.3% carbohydrates [percent of metabolizable energy (ME)], as provided by the manufacturer (Ssniff, Soest, Germany). All diets were supplemented with vitamin and mineral premixes to meet the demands of the National Research Council (NRC) for growing rats.27 For Vitamin D₃, CH contained 1000 IU vitamin D₃/kg diet; LC-HF-1, 1800 IU vitamin D₃/kg diet; and LC-HF-2, 1500 IU vitamin D₃/kg diet. Dietary calcium (Ca) was determined by flame-emission spectography, and dietary phosphorus (P) was determined photometrically with ammonium molybdate and ammonium vanadate in HNO₃ (Ca and P contents: CH: 10.7 and 7.9 g/kg; LC-HF-1: 10.5 and 10.0 g/kg; and LC-HF-2: 6.4 and 4.0 g/kg).

Rats then were pair-fed on an isoenergetic basis for 4 weeks. Pair feeding was chosen because rats of the same age consumed about 20% more of the LC-HF diets in previous ad libitum experiments. For the pair-feeding procedure, daily (ad libitum) food intake of the control chow group was measured, which allowed subsequent calculation of daily energy intake (ME). Thus daily amounts of diet allocated to the LC-HF groups were based on the previously calculated energy intake. Rats in LC-HF groups consumed the whole food allocated to them. By this method, all three groups consumed equal amounts of energy each day, which ensured that any effects observed were due to the macronutrient composition of the diet and not due to differences in caloric intake between groups. All procedures were approved by the Upper Bavarian Government's ethical committee for animal experiments.

Dissection of rats

After 4 weeks on the respective diet, rats were given access to food for 1 hour after lights out and then fasted for 6 hours before decapitation under isofluran anesthesia. Body length (nose to rump) was measured before decapitation while anaesthetized rats were in the ventral position. One observer stretched the rats until snout, spinal column, and middle of the pelvis were in a straight line, and another observer measured nose-to-rump length by use of a caliper. Blood was collected from the trunk after decapitation. Serum samples were stored at −80°C until analysis. Rats were dissected, and epididymal and perirenal fat pads of one side were excised, carefully freed from adherent tissues, and weighed to the nearest 0.1 mg (Scaltec Instruments, Goettingen, Germany). Humeri, tibiae, and femura of both sides were removed for further analyses.

Bone analysis

Peripheral quantitative computed tomography (pQCT)

Left tibiae were used for pQCT (n = 8/group) and µCT analysis (n = 4/group). Scans were made at 10% tibial length from proximal end (Stratec Medizintechnik, Pforzheim, Germany). The in-plane pixels were 70 × 70 µm with a section thickness of 500 µm. Threshold for trabecular, subcortical, and cortical bone was set to 280, 322, and 429 mg/cm³, respectively. Using the software provided by the manufacturer, the following parameters were evaluated in each bone compartment (total, cortical and subcortical, trabecular and cortical): bone content (mg/mm), bone density (mg/mm³) and area (mm²).

Micro-computed tomography (µCT)

Of 24 samples, 12 tibiae (4 from each group) were randomly chosen for µCT scans (Scanco Medical, Bruettisellen, Switzerland). A total of 450 slices were scanned from the metaphyses distal to the growth plate. Scans were obtained using 11 µm resolution with 200 ms integration time. Using the software provided by the manufacturer, the following parameters were evaluated: total volume (mm³), bone volume (mm³), bone volume fraction (BV/TV), trabecular number (1/mm), trabecular thickness (mm), trabecular separation (mm), connectivity density (1/mm³), structure model index (SMI), and degree of anisotropy (DA).

Mechanical properties

A three-point bending test was performed using a mechanical testing machine (Zwick, Ulm, Germany). Before testing of each tibia (n = 8/group), the fibula was removed. Span length was set to 18 mm. Tibiae were tested at a constant cross-head speed of 9.25 mm/min to failure in the mediolateral direction. For each test, a load-displacement curve was recorded, and the following machanical parameters were determined: maximum load (N), stiffness (N/mm; determined by the slope of the linear region), and energy to failure (N-mm; determined as the area under the curve until the maixmum load).

Histology of humeri

Humeri were stored in 4% paraformaldehyde immediately after dissection of the rats. The right humeri were decalcificated in Osteosoft (Merck, Germany) for 48 hours, and then from each bone, six to eight 1 mm³ samples from the epiphysis and diaphysis were randomly sampled, postfixed in 1% osmium tetroxide, and routinely embedded in Epon.28 Epon blocks were trimmed with a TM60 Reichert-Jung milling machine (Leica, Germany), and 0.5 µm semithin sections were obtained with a Reichert-Jung Ultracut E (Leica, Germany). Semithin sections were stained with toluidine blue O and safranin. Fat cells are stained in a light-blue to grayish color. To verify fat cells, Sudan black staining was performed. Light microscopic pictures of toluidine blue– and safranin-stained semithin sections were acquired with a Leica DFC 320 camera (Leica, Germany), using a ×25 objective.

Quantitative real-time PCR from bone marrow

Right femura were opened out (n = 8/group), whole bone marrow was extracted, stored immediately in RNA later solution (Applied Biosystems, Darmstadt, Germany), and kept at −80°C until RNA isolation. Isolation of RNA was performed using the SV Total RNA Isolation System (Promega, Mannheim, Germany) following standard procedures. Briefly, 1 µg of RNA was transcribed in one run applying the SuperScript III First-Strand Synthesis SuperMix for qRT-PCR (Invitrogen, Karlsruhe, Germany).

Quantification of mRNA abundance was performed by real-time PCR detection using a Stratagene Mx3000 instrument (Stratagene, La Jolla, CA, USA) and SYBR-green as a double-stranded DNA–specific fluorescent dye (iQ SYBR Green Supermix, BioRad, Munich, Germany). Amplification mixes (25 µL) contained 0.5 µL cDNA solution, 2× SYBR Green PCR Supermix, 0.5 µL of each primer (10 pmol/µL), and nuclease-free water. Amplification primers (Table 1) were taken from the literature or designed using the open-source software Primer3.

Table 1. Primer Sequences, Annealing Temperatures and Sources of Primers Used for qRT-PCR

Gene	Forward primer	Reverse primer	°C	Ref.
Runx2	ACAACCACAGAACCACAAG	TCTCGGTGGCTGGTAGTGA	60	48
Osterix	TGACTGCCTGCCTAGTGTCTACA	TGGATGCCCGCCTTGT	60	49
C/EBPβ	GCC ACG GAC ACC TTC GAG G	CGG CTC CGC CTT GAG CTG	62	50
18S-RNA	GGG AGG TAG TGA CGA AAA ATA ACA AT	TTG CCC TCC AAT GGA TCC T	60	Primer3

PCR runs consisted of an initial denaturation step at 95°C for 5 minutes, 40 cycles consisting of 10 seconds at 95°C, 30 seconds at the respective annealing temperature (see Table 1), and 45 seconds at 72°C followed by a melting curve. Each PCR included triplicates of cDNA for the gene of interest, no template control, and five dilutions of cDNA pooled from all samples for the gene of interest and for the reference gene 18S-rRNA to calculate the corresponding amplification efficiency (E = 10^–(1/b – 1, where b is the regression coefficient). The parameter Ct (cycle threshold) is defined as the cycle number at which fluorescence intensity exceeds a fixed threshold. Relative mRNA expression for each gene of interest (I) was calculated using the formula (1 + E_[I])^–Ct[I]/(1 + E_[18S-RNA])^{–Ct[18S-RNA]}.28 Data were evaluated by MxPro (Stratagene, La Jolla, CA, USA) and Microsoft Excel software. All results were normalized to expression of housekeeping gene 18S-RNA. Controls have been set to 100%, and expression of LC-HF groups is shown as a percentage of chow controls.

Serum analysis

Procollagen type 1 amino-terminal propeptide (P1NP), osteocalcin, CrossLaps, IGF-1, 25-hydroxyvitamin D (all assays by IDS, Boldon, UK), GH (Diagnostic System Laboratories, Webster, PA, USA), leptin, Insulin-like growth factor binding protein 2 (IGFBP-3, Mediagnost, Reutlingen, Germany), and parathyroid hormone (rat intact PTH, Immutopics International, San Clemente, CA, USA) were measured in serum samples (n = 8/group) with commercially available kits as per manufacturers' instructions. Calcium (Ca) and phosphate (P) were measured by an automated system (Cobas Integra 800, Roche Diagnostics, Mannheim, Germany).

Statistical analysis

Statistical analysis was performed using the SPSS software package (SPSS, Inc., Version 15.0, Chicago, IL, USA). For the statistical comparison between dietary groups, ANOVA with subsequent Bonferroni post hoc test was performed, and p values of less than 0.05 were considered significant. A nonparametric Spearman correlation was performed to determine the correlation between IGF-1 or P1NP levels and mechanical strength in rat tibiae. In the figures, means labeled with different letters are significantly different (a versus b or a versus c indicates a statistically significant difference, a versus a or ab means no significant difference). All data are presented as means ± standard deviation (SD).

Results

Body weight and body length during the feeding period

Rats fed with the CH or LC-HF-1 diet constantly gained body weight (BW) and were significantly heavier at the end of the 4-week feeding period than at the beginning of the experiment (Fig. 1). BW gain of LC-HF-1–fed rats tended to be lower (not significant) when compared with CH controls. Rats on the LC-HF-2 diet also increased their BW during the feeding experiment. However, BW gain was significantly lower (p < .001) when compared with rats fed CH or LC-HF-1 (BW gain after 4 weeks: CH: 189.8 ± 15.8 g, LC-HF-1: 170 ± 8 g; LC-HF-2: 66.7 ± 23.1 g; n = 8/group; see Fig. 1). Rats fed LC-HF-1 were slightly shorter and rats fed LC-HF-2 had a clearly shorter body length compared with CH controls (CH versus LC-HF-1: p < .01; CH and LC-HF-1 versus LC-HF-2: p < .001; n = 8/group; Table 2).

Figure 1. Body weight (BW) development in the different groups during the 28-day diet course (n = 8/group). Rats fed CH and LC-HF-1 constantly gained BW. In contrast, rats fed LC-HF-2 had gained BW only in the second half of the dietary intervention. Data are presented as means ± SD.

Table 2. Body and Long Bone Lengths, Absolute and Relative (Weights Normalized for BW) Fat Pad Weights (n = 8/Group)

Measurement parameter/diet group	Chow	LC-HF-1	LC-HF-2
Data are presented as means ± SD. Means with different superscript letters in each row are significantly different (p < .05).
Body length (cm)	23.9 ± 0.7^a	22.7 ± 0.3^b	20.2 ± 0.7^c
Tibia length (mm)	35.9 ± 1.1^a	35.5 ± 0.6^a	32.5 ± 0.6^b
Femur length (mm)	30.6 ± 0.5^a	30.1 ± 0.6^a	28.1 ± 0.3
Humerus length (mm)	25.3 ± 0.6^a	25.5 ± 0.7^a	23.5 ± 0.4^b
Epididymal fat pad weight (absolute) (g)	1.99 ± 0.26^a	3.02 ± 0.46^b	1.75 ± 0.58^a
Epididymal fat pad weight (normalized for BW)	0.63 ± 0.08^a	1.01 ± 0.16^b	0.86 ± 0.19^b
Perirenal fat pad weight (absolute) (g)	1.18 ± 0.19^a	2.33 ± 0.55^b	1.71 ± 0.81^a,b
Perirenal fat pad weight (normalized for BW)	0.37 ± 0.07^a	0.78 ± 0.18^b	0.83 ± 0.3^b

Long bone growth and visceral fat

Humeri, tibiae, and femur bones of the LC-HF-2 group were significantly shorter (CH and LC-HF-1 versus LC-HF-2: p < .001; n = 8/group; see Table 2), whereas no significance difference in long bone lengths was present between the CH and LC-HF-1 groups. Absolute weights of epididymal and perirenal fat pads were significantly higher in the rats fed the LC-HF-1 diet (p < .01; n = 8/group; see Table 2). When fat pad weights were normalized to BW, rats in the LC-HF-2 group also accumulated significantly more fat (p < .001; n = 8/group; see Table 2).

Bone mineral density/bone volume fraction

Measurement of BMD by pQCT revealed a reduced BMD in tibiae of both LC-HF groups (p < .001) compared with CH controls (BMD pQCT: CH: 423 ± 11 mg/cm³, LC-HF-1: 357 ± 11 mg/cm³; LC-HF-2: 352 ± 23 mg/cm³; n = 8/group; p < .01; Fig. 2A). Similarly, µCT analysis of tibiae confirmed reduced bone volume fraction for rats fed the LC-HF-1 and LC-HF-2 diets (BV/TV µCT: CH: 0.103 ± 0.016, LC-HF-1: 0.06 ± 0.003; LC-HF-2: 0.053 ± 0.01; p < .01 and p < .001, respectively; n = 4/group; see Fig. 2B).

Figure 2. (A) Bone mineral density (BMD) in tibiae measured by peripheral quantitative CT (pQCT). In rats fed the LC-HF-1 and LC-HF-2 diets, BMD was reduced significantly (n = 8/group, p < .01). (B) Bone volume fraction (BV/TV) determined in tibiae by µCT. LC-HF groups showed significantly reduced BV/TV (n = 4/group, p < .01 and p < .001). Data are presented as means ± SD. Different letters indicate significant differences between the groups (p < .01).

Bone microstructure

Analysis of bone microstructure by µCT revealed a lower trabecular number in LC-HF groups (CH versus LC-HF-1: p < .01; versus LC-HF-2: p < .001; n = 4/group; Fig. 3A), whereas trabecular thickness showed no difference between the three groups (see Fig. 3B). Trabecular separation was increased significantly with both LC-HF diets (CH: 0.4 ± 0.15 mm; LC-HF-1: 0.95 ± 0.14 mm; LC-HF-2: 1.17 ± 0.17 mm; p < .01). Also, connectivity density was lower in LC-HF groups compared with CH controls. However, statistical significance was reached only when comparing CH and LC-HF-2 (p < .05; see Fig. 3C). Neither SMI nor DA was significantly different between the three groups (SMI: CH: 2.37 ± 0.11; LC-HF-1: 2.39 ± 0.05; LC-HF-2: 2.33 ± 0.09; DA: CH: 2.68 ± 0.1; LC-HF-1: 2.76 ± 0.18; LC-HF-2: 2.76 ± 0.23). Figure 3D illustrates the digital reconstruction of µCT scans representative for each group and shows the reduced trabecular number in both LC-HF groups.

Figure 3. µCT analysis of trabecular number (A), trabecular thickness (B), and connectivity density (C) in tibiae (n = 4/group). Trabecular thickness was not different between the three dietary groups. In contrast, rats fed the LC-HF-1 and LC-HF-2 diets showed significantly reduced trabecular numbers and also reduced connectivity density (significant for LC-HF-2). (D) Digital reconstruction of µCT scans representative for each group (n = 4). Black bars: 1 mm. Data are presented as means ± SD. Different letters indicate significant differences between the groups (p < .1).

Mechanical properties

Three-point-bending tests of the tibial midshaft showed reduced maximum load to failure in LC-HF groups compared with CH controls (maximum strength: CH: 51.7 ± 4.5 N; LC-HF-1: 46.7 ± 5 N; LC-HF-2: 36.6 ± 3.8 N; CH versus LC-HF-1: p = .064; CH and LC-HF-1 versus LC-HF-2: p < .001; n = 8/group; Fig. 4A). Bone stiffness and energy absorbed under the curve to failure were lower with LC-HF-1 (not significant) and significantly reduced with LC-HF-2 compared with CH controls (for stiffness: CH versus LC-HF-2: p < .05; for energy to failure: LC-HF-1 versus LC-HF-2, p < .01, and CH versus LC-HF-2, p < .001; see Fig. 4B, C).

Figure 4. (A) Mechanical strength of tibiae midshaft evaluated by three-point bending tests (n = 8/group). Maximum load to failure was lower in the LC-HF-1 group (p = .064) and significantly lower in the LC-HF-2 group. Data are presented as means ± SD. Different letters indicate significant differences between the groups (p < .001). (B) Bone stiffness (n = 8/group) was significantly lower in the LC-HF-2 group compared CH controls. Data are presented as means ± SD. Different letters indicate significant differences between the groups (p < .5). (C) Energy absorbed under the curve to failure was reduced in the LC-HF-1 group (not significant) and significantly lower in the LC-HF-2 group compared with CH controls. Data are presented as means ± SD. Different letters indicate significant differences between the groups (p < .01). (D, E) Serum IGF-1 or P1NP levels showed a strong positive correlation with maximum load to failure in rat tibiae (n = 24; part D: p < .001, r = 0.84; part E: p = .02, r = 0.47).

Markers of bone turnover, calcium homeostasis, and GH/IGF system

LC-HF groups had about two- to threefold increased leptin levels (Table 3). The bone-formation marker P1NP was reduced significantly in the sera of rats fed the LC-HF-1 and LC-HF-2 diets (p < .05 and p < .001, respectively), whereas no significant differences between the three groups were present for osteocalcin or CrossLaps. PTH, 25-hydroxvitamin D, calcium, and phosphate were similar in all three diet groups (see Table 3).

Table 3. Serum Parameters of Bone Turnover, Calcium, Phosphate, PTH, 25-Hydroxyvitamin D, Leptin, and the GH/IGF System (n = 8/Group)

	CH	LC-HF-1	LC-HF-2
Data are presented as means ± SD. Means with different superscript letters in each row are significantly different (p < .05).
P1NP (ng/mL)	13 ± 2.5^a	10.3 ± 1.3^b	7.7 ± 1.5^c
Osteocalcin (ng/mL)	993 ± 89^a	952 ± 112^a	894 ± 84^a
CrossLaps (ng/mL)	37.1 ± 6.9^a	39 ± 7^a	44.4 ± 10.3^a
Parathyroid hormone (pg/mL)	365 ± 118^a	317 ± 205^a	325 ± 245^a
25(OH2)D (nmol/L)	54.3 ± 8.9^a	47.1 ± 11.3^a	63.2 ± 15.8^a
Calcium (mmol/L)	2.7 ± 0.1^a	2.6 ± 0.1^a	2.8 ± 0.3^a
Phosphate (mg/dL)	10.3 ± 1.2^a	9.3 ± 0.8^a	9.2 ± 1.7^a
Leptin (pg/mL)	278 ± 158^a	857 ± 300^b	735 ± 436^b
GH (ng/mL)	107 ± 84^a	68 ± 91^a	67 ± 78^a
IGF-1 (ng/mL)	1544 ± 251^a	1312 ± 96^b	562 ± 94^c
IGFBP-3 (ng/mL)	391 ± 58^a	291 ± 62^b	144 ± 15^c

Mean random GH was reduced in LC-HF groups. However, the difference failed to reach statistical significance. Serum IGF-1 was reduced in the LC-HF-1 group (p < .05) and was even lower in the LC-HF-2 group (p < .001) compared with CH controls. Also, IGFBP-3 was significantly lower in both LC-HF groups (p < .01 and p < .001). Of note, IGF-1 and P1NP levels correlated strongly with maximum load to failure in tibiae (n = 24, p < .001, r = 0.84 and p = .02, r = 0.47; see Fig. 4D, E).

Histology of bone marrow

To explore the amount of fat cells in bone marrow, we performed a histologic analysis of several bone slides in each dietary group. Histology revealed that rats fed either LC-HF diet (LC-HF-1: Fig. 5B; LC-HF-2: Fig. 5C) exhibited clearly more fat cells in bone marrow (light-blue to grayish cells) compared with CH controls (Fig. 5A). (Figure 5 shows representative images from six to eight stained sections/humerus; n = 3/diet group.)

Figure 5. Histology of bone marrow (humerus). It appears that rats fed the LC-HF-1 and LC-HF-2 diets (B, C) exhibit significantly more fat cells (light-blue to grayish cells) in bone marrow compared with CH controls (A). Semithin sections, toluidine blue O and safranin stain, ×25 objective. Representative images from six to eight stained sections per humerus; n = 3/diet group.

Quantitative real-time PCR

Since changes in P1NP and CrossLaps indicate inhibition of osteoblast activity, we analyzed gene expression of markers of osteoblastogenesis. C/EBPβ, Runx2 (mesenchymal), and osterix (preosteoblast) transcription factors were downregulated by 70% to 80% in LC-HF–fed rats compared CH controls (n = 8/group; p < .01 for Runx2 and osterix and p < .05 for C/EBPβ; Fig. 6A–C).

Figure 6. Expression of Runx2, osterix, and C/EBPβ transcription factors in bone marrow by quantitative real-time PCR. All results were normalized to expression of housekeeping gene 18S-RNA. Controls were set to 100%, and expression in the LC-HF groups is shown in comparison with CH controls (p < .5 for C/EBPβ; p < .01 for Runx2; p < .01 for osterix; n = 8/group). Data are presented as means ± SD. Different letters indicate significant differences between the groups.

Discussion

The main finding of our study is that both LC-HF diets have unfavorable effects on bone growth, bone structure, and mechanical properties in rats. Regardless of the dietary protein content in the diets used, pQCT and µCT independently revealed a significant reduction in BMD and bone volume fraction (BV/TV), respectively. In contrast, length of long bones was reduced with the LC-HF-2 diet only, the ketogenic diet with the low protein content. Reduced trabecular number and reduced levels of P1NP but unchanged trabecular thickness and CrossLaps, indicated that the LC-HF–induced reduction of bone volume fraction was a result of reduced bone formation owing to reduced osteoblast activity. In line with a reduced osteoblast activity detected in serum, the expression of transcription factors in bone marrow driving stem cell differentiation toward the osteoblastic lineage also was reduced with both LC-HF diets. These data from an animal model might give functional insights explaining why children suffering from epilepsy and treated with ketogenic LC-HF diets show impairments in longitudinal growth. In addition, one might conclude that use of LC-HF diets in adult patients already suffering from impairments of bone metabolism such as osteoporosis should not be recommended.

LC-HF diets in general are rather extreme diets, designed to consist mainly of crude fat and protein. In our study, both LC-HF diets were supplemented with mineral and vitamin premixes to meet the demands of the NRC for growing rats.27 We investigated serum calcium, phosphate, 25-hydroxyvitamin D, and PTH. Since we did not find any differences between the three groups, we have no indication for changes in calcium metabolism to be causative for the deficit bone structure in rats fed the LC-HF diets. A previous study analyzed calcium requirements of growing rats by adding staggered concentrations of calcium (from 1 to 7 g Ca/kg) to the diet. In that study, a threshold of 2.5 g Ca/kg, less than half the calcium concentration that we used in our extreme LC-HF-2 diet, was proposed to induce modest impairments of bone mass, structure, or biochemical strength.29 In order to clarify the potential impact of low dietary calcium and protein in the LC-HF-2 diet, we previously performed a separate experiment using the same diets but with 16-week-old rats. In that experiment, the apparently absorbed amount of calcium was the same in LC-HF-2 and CH groups and even higher in the LC-HF-1 group (data not shown). Protein digestibility and daily nitrogen balance in 16-week-old rats fed the LC-HF-2 diet have shown that these rats were still able to maintain an overall positive nitrogen balance (data not shown). Thus the low protein content but not a change in calcium absorption with the LC-HF-2 diet seems to be responsible for the reduction in longitudinal growth in this group. To our knowledge, no previous study has investigated bone growth in rats fed LC-HF diets. A recent study in mice fed a moderate LC-HF diet with a carbohydrate content of 7% and a protein content comparable with our LC-HF-1 diet also reported a reduced BMD.30 In that study, the observed reduction in BMD was relatively small and not as prominent as in our rats fed the LC-HF diets. Apart from the potential species differences, it seems that the very low carbohydrate content used in our diets (∼1%) has aggravated loss of BMD and reduction in osteoblast activity.

We have previously reported11 and also were able to show in this study that rats fed both LC-HF diets gained significantly more visceral fat accompanied by increased serum leptin levels. Also, rats fed HF but normal-carbohydrate diets show an obese phenotype. A number of investigations analyzed the relationship between HF diets, increased fat mass, and bone quality in rodents. It has been reported that diets with HF contents, e.g., Western-style diets, impair bone metabolism8 and result in low bone mass, poor bone quality, and decreased bone mineralization and mechanical strength.7, 31–33 A connection between increased fat mass and reduced bone quality might be leptin. Leptin has been suggested to control bone resorption by controlling expression of RANKL and CART (cocaine amphetamine regulated transcript), two factors that are thought to regulate osteoclast differentiation.34 Although discussed controversially, leptin also has been proposed to support bone formation. Leptin treatment has been shown to induce osteoblast formation35 and to increase bone formation in leptin-deficient ob/ob mice.36 In addition, biomechanical properties of cortical bone were not ameliorated in GH-deficient dwarf rats with high leptin levels induced by feeding an HF diet.37 In contrast, in leptin-deficient ob/ob and lean mice, intracerebroventricular (i.c.v.) injections of leptin decreased bone mass,38 and it was suggested that leptin inhibits bone formation by influencing the sympathetic nervous system.39 As a possible explanation for these partly discrepant findings, it has been reported recently that leptin exerts a bimodal dose-dependent effect on bone growth and remodeling. While low-dose administration of leptin prevented tail suspension induced bone loss in rats, high doses of leptin inhibited bone growth and reduced bone mass.40 Because in our study, leptin levels were more than twofold increased in rats fed LC-HF diets, we speculate that in our investigation accumulation of visceral fat and increased leptin levels contributed to the poor bone quality.

Interestingly, leptin treatment has been reported to decrease circulating IGF-1.40 Thus the low IGF-1 levels also could be a result of increased leptin in our rats. The GH/IGF system represents another crucial hormonal system to control bone growth.20 Analysis of body composition and bone mineral density in 43 different inbred mouse strains fed with an HF diet and 30 inbred strains fed with an LF diet revealed that circulating IGF-1 levels had a direct effect on areal BMD.41 Of note, it has been shown in previous rodent studies that dietary manipulations such as caloric restriction or low protein content can result in reduced circulating IGF-1 levels.9, 10 As we have reported previously11 and confirmed in this study, rats fed with both LC-HF diets had lower circulating GH, IGF-1, and IGFBP-3 levels compared with CH controls. Therefore, the diet-induced reduction in GH/IGF system components probably aggravated the observed bone phenotype in this study.

A decrease in bone volume as seen, for example, in osteoporosis is often accompanied by an accumulation of bone marrow fat.42–44 Similarly, in our study rats, fed either LC-HF diet also exhibited more fat cells in bone marrow compared with CH controls. Adipocytes in bone marrow differentiate from a common mesenchymal cell lineage with osteoblasts, chondrocytes, and myoblasts. Shifting differentiation and survival rates from the osteoblastic to the adipocytic lineage thus could result in an altered bone mass.24 Since we found a lower BMD in rats fed LC-HF diets, we analyzed markers of bone formation and osteoblast activity, respectively. In serum, quantification of P1NP provides a way to measure the generation of bone matrix type I collagen, which is a crucial step for bone formation. Measurement of P1NP in humans is one of the most reliable and sensitive indicators of early bone formation.45 However, until recently, only little was known about P1NP in rodents owing to the lack of commercially available assays. We found a significant reduction of P1NP and unchanged concentrations of bone resorption marker CrossLaps in LC-HF groups. This indicates that reduction in bone mass is caused by impairments of bone assembly rather than by increased bone resorption.

In our study, expression of Runx2, osterix, and C/EBPβ was significantly lower in rats fed either of the LC-HF diets. On a molecular level, Runx2, osterix, and C/EBPβ are among the transcription factors that support osteoblast differentiation in bone marrow. Runx2 is essential for bone formation25 and controls, in concert with osterix, the expression of genes encoding the protein constituents of bone, including alkaline phosphatase, type I collagen, osteocalcin, and bone sialoprotein.26 Mice deficient in C/EBPβ show a delayed bone formation and osteoblast differentiation. In addition, it has been suggested that C/EBPβ could act as a bridging protein, able to bind to the osteocalcin promoter, thus forming an active transcription factor complex with Runx2.26 Thus an unbalanced differentiation shift of bone mesenchymal cells away from the osteoblastic lineage toward the adipocyte lineage might explain the defects in bone formation. Of note, it has been reported that GH modulates the transcriptional function of Runx2 in osteoblastic cells by promoting its inhibitory interaction with Stat3β.46 Furthermore, an IGF-1-dependent regulation of osterix expression during osteoblast lineage progression has been reported.47 Thus the decreased levels of GH and IGF-1 in our experiment possibly mediated Runx2 and osterix downregulation, consequently leading to reduced osteoblast differentiation and BMD in rats fed LC-HF diets.

In conclusion, in rats, unfavorable effects on bone growth, BMD, and mechanical strength occurred with both LC-HF diets, whereas reductions in long bone growth were seen in the extremely high-fat (low-protein) LC-HF diet only. Rats fed LC-HF diets accumulated significantly more visceral and bone marrow fat. Increased leptin and decreased IGF-1 and IGFBP-3 independently or in concert potentially mediated the poor bone quality observed. Markers of bone turnover in serum suggest reduced bone formation but unchanged bone resorption in the LC-HF groups. These results are supported by expression data of transcription factors influencing osteoblastogenesis in bone marrow (Runx2, osterix, and C/EBPβ), which were reduced by about 70% to 80%. Thus reduced bone formation with LC-HF diets might be explained by a lower rate of mesenchymal cells differentiating into osteoblasts.

Disclosures

The authors state that they have no conflicts of interest.

Uncited Rrferences

5, 17