Reduction of anabolic signals and alteration of osteoblast nuclear morphology in microgravity

Preparation of Cells

The MC3T3-E1 osteoblast cell line clonally derived from embryonic mouse calvaria [Hiramatsu et al., 1983, was kindly provided to us by Dr. M. Kumegawa (Josai Dental University, Japan). The cell line was maintained at low passage number. Cells were grown in alpha minimal essential medium (αMEM) with 10% fetal calf serum (Hyclone Labs, Inc., Logan, UT) supplemented with 2 mM L-glutamine (Sigma, St. Louis, MO), 25 mM HEPES, and antibiotic–antimycotic solution (100 U penicillin/ml, 0.01 mg streptomycin/ml, 0.25 mg amphotericin B/ml). Cells were grown in a 37°C incubator with 5% CO₂. They were fed three times a week and passaged at confluence. For spaceflight experimental samples, 120,000–200,000 cells were plated onto non-coated, sterile 11 × 22 mm glass coverslips (Thomas Scientific, Swedesboro, NJ), placed in 6-well plates, and grown in 10% serum-containing αMEM overnight. Cell-coated coverslips were then transferred into 2% serum-containing medium for spaceflight in plungerbox units. In accordance with NASA flight rules, the plungerbox units were held for 17 h in the Shuttle at mid-deck temperature prior to launch. This, combined with low serum-containing medium, placed the osteoblast in a quiescent state prior to Space Shuttle launch. In order to achieve our goal to study changes in nuclear morphology in microgravity, we launched cells in a quiescent condition. Osteoblasts were sera activated on-orbit and µg or 1 g samples collected in-flight (Fig. 1). There were four samples (n = 4) for each time point in µg, 1 g flight, and gr samples. NASA's LSLE (Life Sciences Laboratory Equipment) freezer, was used for sample storage on the Space Shuttle. Supplements, such as dexamethasone β-glycerol phosphate and ascorbic acid were not added to the media since these agents are known to directly effect gene expression and cell morphology of the osteoblast.

Biorack Facility and Osteo Hardware

Cells were grown in the Biorack, a multiuser facility, which consists of incubators with variable gravity centrifuges, a cooler, a freezer, and a sealed glovebox. Two identical Biorack modules were used: one remained on earth (gr samples) and the other was integrated into SpaceHab and flown on each of the Space Shuttle flights STS-76, STS-81, and STS-84. Biorack has an important advantage over other microgravity facilities in that it provides a small radius (78 mm), slow rotating (107.0 ± 0.5 rpm) centrifuge for artificial 1 g in orbit. Because of their proximity, both the µg and 1 g samples experience identical launch vibrations, accelerations, cosmic radiation, and other unknown conditions of flight. In addition, an identical experiment was performed in the Biorack module on earth gr samples with a 2 h delay from in-flight procedures in order to control for any off-nominal operations during the flight. The Osteo (the name of this flight experiment) spaceflight hardware was designed according to European Space Agency (ESA) specifications for use in the Biorack facility and constructed by Centrum voor Constructie Mechatronica (CCM; Neuenen, Netherlands). The hardware consisted of the CCM plungerbox and its Type I container developed for spaceflight cell culture. The Type I container provided a second level of fluid containment. The plungerboxes were designed to provide a sterile environment for cell growth activation and fixation in a microgravity condition. The plungerbox is composed of two independent culture chambers that each hold two 11 mm × 22 mm glass coverslips. Each culture chamber was a separate sample. For each condition a sample size of n = 4 independent cultures was used. Each separate culture chamber has a series of reservoirs filled by either 10% serum-containing αMEM or fixative, which can be transferred into and out of the cell culture compartment by manually releasing a spring-loaded plunger.

Experiment Time Line

Quiescent cells were stored in the Space Shuttle mid-deck locker until 19 h after launch. The astronauts then transferred the samples into a 37°C incubator for 1 h before stimulating the cells to grow in the 1g centrifuge or µg static rack by changing their media from 2% serum to 10% serum-containing medium (T = 0). µg and 1 g samples were fixed during flight at 24 h with 3.7% formaldehyde solution and were stored at +5°C for the remainder of the Space Shuttle mission, while the GITC samples were frozen until return. There were four separate independent samples for each gravity condition at each time point (Fig. 1).

FGF-2 Protein Analysis

Analysis of FGF-2 was completed with a Quantities ELISA Immunoassay (R&D Systems, Inc., Minneapolis, MN) according to manufacturer's instructions. Cellular protein was collected from the GITC subnatant as previously described [Fitzgerald and Hughes-Fulford, 1999.

Analysis of Gene Expression

Table I shows the overview of gene expression measured in these experiments. RNA was extracted and purified by a modified acid guanidium thiocyanate/phenol/chloroform extraction method previously described [Fitzgerald and Hughes-Fulford, 1996; Barry et al., 1999; Fitzgerald and Hughes-Fulford, 1999.

RT-PCR Analysis

The semi-quantitative RT-PCR analysis and primer sequences for many of the genes have been described [Fitzgerald and Hughes-Fulford, 1996; Barry et al., 1999; Fitzgerald and Hughes-Fulford, 1999. To correct for small variations between experiments, each PCR product was compared to an internal standard of cyclophilin or 18 S products derived from the same RT reaction. Since sample size was small (200,000 cells), RNA content was held constant and linear RT-PCR was accomplished by varying the number of PCR cycles. Primers were designed in this laboratory to span introns to eliminate the possibility of detection of DNA. PCR conditions were established so that amplification reaction was stopped in the linear range. Primers not previously described are 18s F-TCA AGA ACG AAA GTC GGA GG; R-GGA CAT CTA AGG GCA TCA CA; CPHI F-CGT CTC CTT TGAG CTG TTT GCA GAC R-CAT AAT CAT AAA C TAA CTC TGC AAT CCA GC; fgf-2 F-AAC TAC AAC TTC AAG CAG AAG AGA GAR-TTA AGA TCA GCT CTT AGC AGA CAT; bcl2 F ACTTGTG GCCCA GATAGGCACCCAG; R CGACTTC GCCGAGATG TCCAGC CAG; IL-1β F-GTC TCT GAA TCA GAA ATC CTT CTA TC R-ATG TCA AAT TTC ACT GCT TCA TCC; cpla2 F-GGA TTC TCT GGT GTG ATG AAG G, R-CCC AAT CTG CAA ACA TCA GC ep1 F-CTG GAG GGG CGT GTC ATT TC, R- GCT GCG GAG GGT GGC TGT GG ep2 F-CTA CGC CGC CTT CTC TTA CAT G, R- GAT GTC TTT CAC CAC GTT TGG C; ep3 CAT CAC CAT GAT GGT CAC TGG C, R-GTC TTC ATG TGG CTG GGA TAC C; LDLr F-CAA TGT CTC ACC AAG CTC T, R-TCT GTC TCG AGG GGT AGC T; cyclin A F AAA CAG CCT GCC TTC ACC ATT CA R-ATA TTC TTC TCC CAC CTC AAC CA; cyclin E, F-CAA GTG GCC TAT GTC AAC GA R- ACT GCT GGG TGG GGG TGT CA; PCNA F-CAC CAA ATC AAG AGA AAG TT, R-GAG AAT GGA GAG AAA AGA AA; c-myc F-CCA ACA GGA ACT ATG ACC TCG, R-CCA CAT ACA GTC CTG GAT G; cjun F-GGA AAC GAC CTT CTA TGA CGA TGC CCT CAA, R-GAA CCC CTC CTF CTC ATC TGT CAC GTT CTT; collagen, F-GTC CCC CTG GCT CTG CTG GTT, R- TTT GGG TTG TTC GTC TGT TTC; β1 Integrin, TGT TCA GTG CAG CCT TCA, R-CCT CAT ACT TCG GAT TGA CC; bax F-CAG CTC TGA GCA GAT CAT GAA GAC A, R-GCC CAT CTT CTT CCA GAT GGT GAG C.

Cell Morphology

Osteoblasts were fixed on coverslips with 3.7% formaldehyde in phosphate buffered saline during spaceflight and stored at 5°C for 1–2 weeks before staining. There was no visible difference between control samples that were immediately stained and samples stored in formaldehyde at 5°C. For nuclear visualization, coverslips were stained with Hoechst dye number 33258 and rinsed in distilled water. The dried osteoblast coverslips were mounted onto slides and photographed with a Zeiss Axioscope microscope at 40× and 100× magnification. Slides were processed at the same time under identical conditions. Photographic exposure was taken at identical exposure times and conditions (Tables VI and VII; Figs. 4-7, 9).

Image Acquisition

Table II shows number of images at the total number of cells evaluated. Images were digitized from the color film slides with an Epson Expression 836XL scanner into Adobe Photoshop (Adobe Systems, Inc., Mountain View, CA) (Table III and Fig. 4). Images were stored as RGB (red, green, blue) TIF format image files (Fig. 5). From these RGB images (Fig. 5a) only the RED channel image (Fig. 5R) was used since the nuclear structure is best represented by this channel.

Image Preprocessing and Segmentation

All image analysis was performed with IDL software (Interactive Data Language, Research System, Inc., Boulder, CO). For segmentation, a global luminance value was automatically estimated from every image L and used for thresholding of the image content into background and rough cell nuclei mask M_r (Fig. 6). Cell cluster consisting of touching or occluding cells were delineated by manual interaction. Border touching cells were deleted resulting in mask M. The cutting lines for the edge of touching cells were set in the center of the overlapping or occluding, typically lens-shaped areas. Nuclei with large overlap with others were rejected from further analysis. For feature extraction, the RED channel, image R (Fig. 5R) was transformed to the fluorescence density image F (Fig. 5c) by the formula:

This is analogous to the well-known transformation from transmitted light T to extinction E. Extinction is equivalent to the optical density in quantitative microscopy.

Feature Extraction

From each cell nucleus, represented by its mask M and the corresponding region in the fluorescent density image F, numerous quantitative features were calculated describing shape, photometric, and textural properties. The latter are understood as properties of the arrangement of fluorescent particles, which in this case is DNA, visualized by Hoechst 33528. These features are described in detail elsewhere [Rodenacker and Bengtsson, 2003 (Fig. 7). For classification, selected features are briefly described in Tables VI and VII. They are illustrated by image sequences showing the changing feature value with appropriate cell images.

Statistics and Classification

All statistical evaluations were completed using SAS (SAS Institute, Inc., Cary, NC) and BMDP (Statistical Software, Inc., Los Angeles, CA) program packages. All cells from specimens belonging to the same experimental conditions were pooled and stepwise linear discriminant analysis was applied. From the whole feature set only those features were used in the selection steps that were considered independent of possible global acquisition influences such as relative features and which were univariate significant. Typically, features measuring area or total fluorescent intensity were only used after normalization, for example, particle area divided by nuclear area. Due to the small number of cells, only up to seven (three-class) or six (two-class) features respectively were stepwise selected on the basis of F-statistics to avoid over fitting. The value for the first chosen feature is univariate whereas the following F-values are multivariate reflecting the impact of results after using this feature together with the already selected features. The classification matrices show the results of the hold-one-out (jackknifed) method because no test set was available. All statistical evaluations were done at the 95% confidence level. For display of the three-class discrimination, the canonical variables are used which are linear combinations of the selected features. Linear discriminant analysis results in the so called discriminant function defined by a ((n − 1)-classes × m-feature) matrix D. The canonical variables CAN are than defined as

This is the product of the discriminant matrix D and the selected features. This is a method to project the observations for visualization.

Gene Expression and Protein Levels in Microgravity

We analyzed induction of 24 osteoblast genes using linear semi-quantitative RT-PCR. We found reduced gene expression in 10 of these genes; no change in 9 of the genes and copy number was too low to measure in the remaining genes examined. Osteoblasts were activated during space flight in µg with and without a 1 g field and results compared to gr samples. Induction of PCNA, TGFβ, cox-2, cpla2, osteocalcin, c-myc, fgf-2, bcl2, and bax were significantly depressed in µg when compared to gr (Figs. 2 and 3). Artificial onboard gravity was able to normalize the induction of c-myc, cox-2, TGFβ, fgf-2, bax, and bcl2 as well as FGF-2 protein synthesis in spaceflight samples.

There was no change in quiescent osteoblast PGE₂ levels (gr 415 ± 163 vs. µg 453 ± 189). Even though the cox-2 message was reduced at 24 h, the PGE₂ levels were high. However, in-flight at 24 h, both the 1 g and the µg samples had a significant rise in PGE₂ levels: 1 g 6,065 ± 870 and µg 6,959 ± 1132 versus gr 3,904 ± 1673 (P < 0.005).

Since it has been previously demonstrated that a lack of mitogen causes elongation of nuclei [Foote et al., 1996, we measured endogenous levels of cellular fgf-2 message and protein in these microgravity grown samples. In our studies, we found that the fgf-2 message induction was depressed 24 h after activation (Fig. 3). Moreover, when the cellular protein levels were measured in these cells, we found that FGF-2 protein synthesis was also significantly reduced in µg flight, however, the presence of artificial gravity in-flight increased the ability of the cells to synthesize fgf-2 (Fig. 3) 24 h after serum activation. Finally, changes in bcl2 message (Fig. 3) induction parallel that of fgf-2 message. The bcl2 message was significantly decreased in microgravity compared to ground and 1 g samples.

Three-Class Case

This results in the most discriminative features P2A, HETERO, GCV, RHA, RHUA, NR1, and HLM3. A short description of the features selected is given in Tables VI and VII. Cell nuclei show clear differences in shape and texture. The first selected feature (P2A) is shape, remaining features are all texture related. Variation of P2A is based on the deviation from circularity to ellipticity. Further differences are reflected by several features sensitive to the different arrangement of bright fluorescent particles (HETERO) inside the darker nucleus surrounding and the less pronounced change between background inside the nucleus and bright particle (GCV).

The classification result is listed in Table III and illustrated in Figure 8 using the 1st and 2nd canonical variables (CAN1, CAN2) with the class discrimination lines. Using three-class discrimination analysis, it is clear that the gr and µg cell populations are significantly different from one another. The first canonical variable CAN1 alone permits a nearly complete discrimination at about CAN1 = 0.8 of the µg cells from the others (see Fig. 8). A sequence of cell images is displayed in Figure 9 using the first canonical variable as order criteria. Note the change of particle structure from left to right.

Two-Class Cases

They were analyzed to find out more about the changes from one experimental condition to another in terms of most selective features.

From ground (gr) to 1 g (in-flight) (Table IV): The selected features HETERO, ELNENL, RHA, and NR4 represent pure textural changes.

From 1 g (in-flight) to µg (Table VA): The sequence of selected features starts with P2A, a shape factor showing the influence of gravity on growth form, followed by texture features HLM3 and RHUA.

From ground (gr) to µg (Table VB): For completeness the direct transition from ground (GR) to µg is also outlined. Chosen features are HETERO, RHUA, and RATIO. The first two are pure textural and the last is a shape feature similar to P2A.

Table VI. Most Important Features From Three-Class Discrimination With Image Sequences of Increasing Feature Values

Table VII. Most Important Features From Two-Class Discriminations With Image Sequences of Increasing Feature Values

The two-class discrimination analysis provided a comparable result with the heterogeneity of nuclear texture (HETERO) showing the greatest difference between gr and µg (correct class 94.7%). A similar difference, with a lower correlation was observed for 1 g and µg (correct class 86.4%). Furthermore, the 1 g cells have very similar morphology to the ground cells (correct class 78.7%). In other words, the ground and 1 g samples are more similar to each other than they are to µg cells. The morphological deviation of µg cells may be due to the influence of gravity on the arrangement of nuclear particles as a result of altered gene expression. All other selected features (HLM3, RHUA, NR2, NC9) except the last are features of particle arrangement. RATIO reflects an additional shape influence.