Synovial tissue was obtained from RA and OA patients during knee replacement surgery as described earlier 24. Control synovial tissue samples were obtained from patients with joint trauma during routine arthroscopy or open joint surgery for diagnostic and therapeutic procedures. All patients were informed of the purpose of the study and gave written consent. The study was approved by the Ethics Committee of the University of Regensburg. Characteristics of patients are given in Table 1.

The tissue preparation for histological studies was performed as previously described 24. Cryosections (5 μm) of at least three different formaldehyde-fixed synovial tissue samples from each patient/control were used. We used primary antibodies against the clock proteins BMAL-1 (1:100, polyclonal, rabbit versus human, Dianova, Hamburg, Germany), CLOCK (1:100, polyclonal, rabbit versus human, Calbiochem, Darmstadt, Germany), Period 1 (1:100, polyclonal, rabbit versus human, Millipore, Eschborn, Germany), Period 2 (1:100, polyclonal, rabbit versus human, Biozol, Eching, Germany). The secondary antibody was coupled to the fluorescent dye Alexa Fluor 546 (polyclonal, goat versus rabbit, Invitrogen, Karlsruhe, Germany). Non-specific binding sites were blocked with phosphate-buffered saline containing 10% fetal bovine serum, 10% bovine serum albumin, and 10% normal chicken serum or 10% normal goat serum for 45 minutes at room temperature. The samples were then incubated with the respective primary antibody for 3 hours at room temperature, washed and then incubated with specific secondary antibody for 90 minutes. After 4'-6-diamidino-2-phenylindole (DAPI; Roche, Mannheim, Germany) staining, slides were covered with fluorescence mounting medium (DAKO, Hamburg, Germany) and stored at 4°C until microscopy (performed within four days). Control staining was carried out with the secondary antibody alone or using unspecific rabbit serum instead of the primary antibody. Control staining always yielded a negative result.

Mixed synoviocytes were isolated from synovial tissue as described before 24. The cells were cultured over three to six passages yielding a homogenous cell population of synovial fibroblasts. In an earlier study 25, it was demonstrated that only after seven to eight passages more than 10% of the genes were differentially expressed. Thus, passage 6 was used as the upper limit. The same study also demonstrated that doubling rate was constant for up to five passages and decreased after passages 6 to 8 25. With this information in mind, we focused on passages 3 to 6, which was necessary due to the enormous amount of cells needed.

Preliminary experiments did not show a difference in molecular clock mRNA in synovial fibroblasts of passage 3 to 6. Per well, 150,000 cells were cultured in culture medium (RPMI 1640, Sigma, Steinheim, Germany) containing 10% fetal bovine serum, 1% penicillin/streptomycin and 0.1% amphotericin B for 24 hours. Then, cells were starved for 24 hours in medium without fetal bovine serum in order to synchronize cellular rhythms as previously reported 26. To start daily clock rhythms, cells received a serum shock (10% fetal bovine serum) in culture medium described above. At time t = 0, cells were treated without additional cytokines (control) or with IL-1β or TNF. Starting at t = 0, cells were harvested every 2 hours for a period of 72 hours (this is a high time resolution in these types of experiments). Harvested cells were washed and stored in RNA later solution (Sigma, Steinheim, Germany) at 4°C and then at -30°C until RNA isolation.

Some of the experiments were repeated with the MH7A synovial fibroblast cell line. This cell line was derived from immortalized synovial fibroblasts of a patient with RA using the SV40 T antigen 27. MH7A were purchased from Riken, Japan, with informed consent of Central Research Laboratories, Kissei Pharmaceutical Co., Ltd., Nagano, Japan. Similar as mentioned for primary synovial fibroblasts, synchronization was established by serum shock. Cells were incubated at indicated concentrations of IL-1β and TNF. RNA isolation was performed as described.

Since most studies on molecular clocks studied mRNA and because protein determination with Western blot was extremely time-consuming (preliminary studies), we used quantitative PCR to detect levels of molecular clock mRNA. RNA was isolated according to the manufacturer's instructions (Nucleo Spin RNA II Kit, Macherey Nagel, Düren, Germany). cDNA was converted from total RNA (RevertAid First Strand cDNA Synthesis Kit; Fermentas, St. Leon-Roth, Germany). For quantitative PCR (qPCR), 1 μl of cDNA preparation, 1 μl of specific primer (100 pmol/μl; Eurofins MWG Operon, Ebersberg, Germany) and Brilliant II SYBR Green qPCR Master Mix (Agilent Technologies, Waldbronn, Germany) were applied in a total volume of 10 μl. The PCR reaction was evaluated by melting curve analysis according to the manufacturer's instructions (Thermal Cycler 7900 HAT, Applied Biosystems, Darmstadt, Germany). Each quantitative PCR was performed in triplicate.

The following primers were used: BMAL-1 forward ATCAGACGATGAATTGAAACAC, BMAL-1 reverse TCATTCTGGCTGTAGTTGAGGA; CLOCK forward ACCCTTCCTCAACACCAAC, CLOCK reverse GACTGGGAATTTATGGACTGAC; Period 1 forward AGTTCCATTGCCTACAGCC, Period 1 reverse GAAGTGCTGTCATGAGTTCC; Period 2 forward GCATTTCATTAAGTGCGTCC, Period 2 reverse GCTTCTCTCTGTCCTCCTTC; Per3 forward GTTGTCGCCATCGTTTTTGCC, Per3 reverse GCTTTGTGCCTCCCACTTTTCC; Cryptochrome 1 forward GGATTGATGCCATCATGACAC, Cryptochrome 1 reverse CCTTCATTCCTTCTTCCCAAC; Cryptochrome 2 forward GCGCTGCGTTTACATTCTC, Cryptochrome 2 reverse CTTGTGTCCAAATCTTCCAGAG; Rev-erbα forward CCCTTCTTCCTCATCTTCCTC, Rev-erbα reverse GATGTTGCTGGTGCTCTTG.

Furthermore, Cdc2 was detected in order to estimate the interrelation to the cell cycle because this factor is detrimental for cell cycle progression 28. The following primers were used for Cdc2: forward ACTGGCTGATTTTGGCCTTGC, reverse AGTTGAGTAACGAGCTGACCCC. The housekeeper 18s was used as control using the following primers: forward CGGCTACCACATCCAAGGAA, reverse GCTGGAATTACCGCGGCTGC. Each gene was normalized to the house keeper mRNA of 18s. The ratio of the mRNA of a specific gene divided by the mRNA of 18s at t = 0 was defined as 1.00 and ratios of every following time point were calculated in relation to t = 0.

Medians of groups were compared by Mann-Whitney U test and Box Plots were demonstrated to respect the non-normal distribution of data (Sigma Plot 11.0, Systat Software, Inc., Erkrath, Germany). For time series analysis, cubic regression was used for the interrelation of time of day and density of cells positive for molecular clocks in synovial tissue (Sigma Plot). In addition, autocorrelation analyses were used to identify a possible period of the rhythm (PASW Statistics, 18.0, IBM SPSS via IBM Germany, Ehningen, Germany). We did not use a sine or cosine wave fitting model because it was unclear whether or not oscillations really exist. Trigonometric wave fitting always yields a rhythm and, thus, it is not useful to apply this technique in a situation where the presence of oscillation is not really known. Elapsed time between serum shock synchronization and first peak of molecular clock mRNA curves was determined by inspection of mRNA curves in every individual. The significance level was P < 0.05.

The presence of molecular clocks is a prerequisite for endogenous cellular rhythms. This prompted us to study protein abundance in primary material of patients with OA, RA and trauma controls. Immunofluorescence demonstrated ubiquitous cellular distribution of BMAL-1, CLOCK, Period 1 and Period 2 proteins in patients with RA and OA (Figure 1). Without variation, all cells demonstrated these four clock proteins. In addition, specimens of trauma controls similarly demonstrated the four molecular clocks (data not shown).

In order to study a possible difference between OA and RA material, cellular density of synovial cells positive for molecular clocks was investigated. While synovial density of positive cells for CLOCK and Period 2 were similar between disease groups, density of BMAL-1-positive cells was higher in RA compared to OA (Figure 2A). For Period 1-positive cells, a similar trend existed (Figure 2A). This indicates that RA synovial tissue provides an environment for increased molecular clock levels.

Since BMAL-1 and Period 1 undulate during the day, we hypothesized that this oscillation can be visible when taking patient material at different operation time points during the day. We had access to material taken between 8:00 am and 4:00 pm of a large number of patients with RA and OA as well as trauma controls. Although one would have expected a difference between morning and afternoon, no statistically significant variation of synovial density of positive cells was observed in these patient groups (Figure 2B).

In preliminary experiments, we evaluated the use of Western blots for further protein analyses of molecular clocks (data not shown). However, due to the time series with 36 time points of cell removal and lack of adequate amounts of primary cells, this technique was abandoned.

Typically, quantitative PCR is used to study undulation of molecular clock mRNA over three days 15, which was also employed in the present study. After serum shock synchronization, synovial fibroblasts start with a huge rise of molecular clock mRNA as demonstrated in Figure 3. The rise of mRNA is immediate for Period 1 and Period 2 and a little delayed for BMAL-1 in both OA and RA patients (Figure 3). Using the median of mRNA of all patients in a disease group at a single time point, we were able to demonstrate oscillation curves with a total of maximal four peaks (Figure 3, yellow areas). However, time series analysis with autocorrelation did not reveal a regular pattern to determine a period of oscillation in OA and RA patients (data not shown). In other words, oscillation is masked by a low signal-to-noise ratio in OA and RA patients.

Only in fibroblasts of healthy subjects did we find a regular rhythm for BMAL-1 and Period 1 as demonstrated by autocorrelation analyses (Additional file 1, Figure S1A). These analyses revealed a period of 24 hours for both molecular clocks (Additional file 1, Figure S1A; time between red coefficients outside the confidence limits).

Since time series did not reveal molecular clock mRNA oscillation in OA and RA patients, we used the average of all 36 time points of every patient for further analyses. IL-1β and TNF are abundant proinflammatory cytokines found in patients with OA and RA. Thus, these two cytokines were studied in order to investigate the influence on molecular clock mRNA. While IL-1β had no influence on average mRNA of BMAL-1, CLOCK, Period 1 and Period 2 in OA synovial fibroblasts (Figure 4A), TNF inhibited mRNA of CLOCK, Period 1 and Period 2 (Figure 4A). This is demonstrated in the oscillation curve of Period 1 mRNA (compare green line of control conditions with brown line in Figure 4B, right panel). There are no influences on the BMAL-1 mRNA oscillation curve (Figure 4B, left panel). Similar to OA synovial fibroblasts, fibroblasts of healthy controls show a slight reduction of average CLOCK mRNA by TNF (Additional file 1, Figure S1B).

In contrast to OA synovial fibroblasts, both cytokines stimulated average mRNA of RA synovial fibroblasts (Figure 5A), which is exemplified by BMAL-1 and Period 1 mRNA oscillation curves (Figure 5B). The different protein levels demonstrated in Figure 2A supports that increased mRNA levels translates into increased protein levels.

In order to intensify these investigations, the RA synovial fibroblast cell line MH7A was used (a large number of cells was obtained from these non-primary cells). Both, IL-1β and TNF dose-dependently increased mRNA of Period 1 and Period 2 (Figure 6A). This was similar for CLOCK mRNA but was not statistically significant for BMAL-1 mRNA (Additional file 2, Figure S2).

In order to investigate a possible cell cycle-dependent abundance of molecular clocks, the ubiquitous cell cycle progression factor Cdc2 was investigated. The oscillation curve of Cdc2 mRNA had one peak at approximately 30 hours after synchronization indicating one concerted proliferation event (Additional file 3, Figure S3A), which was largely different from the immediate peak of molecular clocks (compare Figure 3). Thus, there seems to be no direct interrelation between cell cycle and molecular clocks because no parallel or anti-parallel rise or fall of mRNA was observed. Interestingly, IL-1β and TNF increased Cdc2 mRNA in RA but not in OA or control subjects (Additional file 3, Figure S3B). In the direct comparison of RA and OA oscillation of Cdc2 mRNA, the peak was little delayed in RA compared to OA (Additional file 4, Figure S4, black letters).

Proinflammatory cytokines might influence oscillation characteristics of molecular clocks. As mentioned above, the period of the oscillation cannot be studied due to the rapid loss of rhythmicity (see all curves in Figure 3). Thus, elapsed time from start of synchronization to the first peak of the undulation curve was scrutinized. Both, IL-1β and TNF delayed the first peak of BMAL-1 mRNA oscillation in RA synovial fibroblasts, which did not reach the significance level in OA synovial fibroblasts (Figure 6B). IL-1β and TNF did not change oscillation characteristics of other molecular clocks investigated (Additional file 4, Figure S4).

Obviously, there exist four different groups of molecular clocks as determined by the elapsed time from start to first peak (Additional file 4, Figure S4). One group has the peak between 0 and 6 h (blue), one group between 8 and 14 h (pink), another group between 14 and 17 h (red), and a fourth group between 18 and 19 h (green). Molecular clocks in these different groups reflect the different partners of the intracellular molecular clock network (Additional file 5, Figure S5). However, conditions in RA and OA synovial fibroblasts prevent one or two full oscillation cycles, which are visible in healthy fibroblasts (Additional file 1, Figure S1A).