A) Diamagnetic levitation of Drosophila melanogaster
In the 0 g
* tube, we observed flies levitating freely (i.e. not in contact with any surface, or flying) within 1-2 mm of the levitation point of liquid water (Additional file 2
, and [27
]). This is not unexpected because the flies have a high water content. The net effective weight of a freely levitating fly is zero, in the sense that there is no net (gravitational plus magnetic) force on the fly. Although we observed a few flies levitating freely, the majority remained in contact with the walls, floor and ceiling of the arena enclosed within the tube.
The net effective weight of an individual fly on the walls, floor or ceiling depends on its position within the arena (Figure ). The effective weight is less than 5% of its weight outside the magnet throughout the arena.
We observed that the levitation position of the freely levitating flies varied with their hydration. Dehydrated (dead) flies levitated a few millimetres lower in the magnet than living flies. There was a 1-2 mm difference in levitation position between each of the Drosophila
stages. This is consistent with the greater water content of the embryos and early larvae (more than 80% of total mass), reaching the lower level at late pupae (less than 70% of total mass) in comparison with an average 75% of water content in adults [36
B) Delay of development due to the magnetic field
The results of the "long-term" (22-day) experiments demonstrated that 1-12 hour old embryos can develop fully, progressing from larvae to pupae to imagoes, both in the RPM and in a strong magnetic field up to 16.5 T. Development in the magnet (0 g*, 1 g* and 2 g*) was slightly but reproducibly delayed by one day, compared to the 1 g control outside the magnet, suggesting that metamorphosis can be delayed in one or more developmental checkpoints. A less evident delay in development was observed in the RPM.
Table shows the number of flies that developed from eggs laid in the magnetic field in a "short term" (1-day) experiment. After the females were mated, 25 males and 25 females were selected randomly and placed together in the same container in the magnet for 26 hours at 14°C. The eggs laid during those 26 hours were incubated outside the magnet and the flies that developed from the eggs were counted. The results show that exposure to the strong magnetic field during oogenesis and laying caused a large reduction in the number of adult flies that developed from the eggs. The number of flies that developed from eggs laid in the 16.5 T magnetic field (1 g* tube) was just 31% of the number that developed from eggs laid in the 1 g control tube outside the magnet. In the 0 g* and 2 g* tubes, where there was a significant field gradient, the reduction in the number of flies was even greater, being only 5-6% of the number resulting from the 1 g control.
Imagoes developed from eggs laid during the medium term experiment
C) Magnetic field affects gene expression
Using Affymetrix whole genome microarrays (Drosophila version 2.0 with 18952 probesets and GeneSpring GX), we analysed the gene expression profile of Drosophila exposed to the magnetic field, and compared the results with Drosophila in a temperature-controlled incubator placed well away from the magnet (1 g). We also compared the results of "short-term", "medium-term" and "long-term" experiments. In most experiments we performed three replicates, except in a small number of cases (identified by dashes in Table ) where we were unable to obtain one or more replicates owing to random contamination of the extractions, or time constraints on the use of the magnet. We also compared microarray results from the "long-term" magnet experiment with 5 microarrays from a RPM microgravity simulation experiment. The transcriptional profiles of each experiment are shown in Figure . Here, the profiles have been presented as a "condition tree", in which similarities between transcriptional profiles are reflected in the grouping of the experiments within the tree. For example, the transcriptional profiles from the 0 g*, 1 g* and 2 g* tubes in the "long-term" experiment display readily identifiable similarities and so are grouped together in the diagram. Likewise, common features can also be identified in the profiles from the "medium-term" and "short term" experiments. There were significant differences in the profiles from experiments of different duration. For example, there were clear differences between the profile of flies in a 0 g* tube in a long-term experiment and the profile of flies in a 0 g* tube in a short-term experiment. The variation between experiments of different durations was smaller for females than for males. These results showed that the precise biological state of the organism (i.e. age, gender, temperature) was important in the magnetic field effect.
Figure 2 Clustering of the 22 analysed transcriptional profiles. Clustering of the transcriptional profiles is revealed by a condition tree calculated with a hierarchical cluster algorithm, using Pearson absolute distance metric and the average linkage rule. The (more ...)
The Venn diagram in Additional file 1
, Figure S4 shows the number of genes that changed their expression levels in the magnet compared with the 1 g
external controls located well away from the magnet. The number
of genes affected was nearly independent of the duration of the experiment, or the location (0 g
*, 1 g
* or 2 g
* arena) inside the magnet. Interestingly, the group of genes affected in the long-term experiment was different from the group affected by the medium-term experiment. Similarly, the group of genes affected by the short-term experiment was different from both the long-term and medium term experiments, although there was some overlap in the three groups of genes affected. We identified 496 genes that were sensitive to the strong magnetic field in males, i.e
. genes that are up- or down-regulated in one or more of the long-, medium- and short-term experiments in the magnetic field. Of this group of 496 genes, 105 were common to two different experiments. We found only one gene common to all three experiments (long, medium and short-term), namely CG33070-RB, Sex Lethal, encoding an RNA binding protein.
In females, 474 genes were sensitive to the magnetic field, of which 115 were common to flies in two different experiments. Fourteen genes were common to flies in all three experiments; three of these genes have been annotated as heat shock proteins.
Less than 10% of the magnetic field-sensitive genes were common to both males and females; 5 were observed in two or more experiments, with 47 observed in one or more experiments.
We also analysed the short-term, medium-term and long-term experiments separately, in order to identify the differentially-expressed genes induced by the magnetic field. The number
of genes sensitive to the magnetic field in the individual experiments is shown in a series of Venn diagrams in Additional file 1
, Figure S4.
D) Isolating the effect of magnetically-altered effective weight from other effects of the strong magnetic field
The above results indicate that the strong magnetic field present in all three tubes (0 g*, 1 g* and 2 g*) had a significant effect on gene expression of the flies. In order to locate genes that could play a role in gravisensing or adaption, it was necessary to account for the effects of the strong magnetic field. We attempted to isolate the effect of the vertical diamagnetic force on flies (which alters the effective net weight of the flies), from any other effects of the magnetic field by comparing the gene response of the flies in the 0 g* and 2 g* tubes to those in the 1 g* tube
Two different approaches were used:
• Approach 1: A list of genes that were up- or down-regulated in the 0 g* tube in the magnet was compiled and compared with the external control outside the magnet (1 g). We repeated this for the 1 g* and 2 g* tubes, and then removed from the 0 g* vs 1 g and 2 g* vs 1 g lists those genes that appeared in the 1 g* vs 1 g list.
This approach was based on the results in Additional file 1
, Figure S4, first and second rows, in which it is evident that few genes were common to flies in all tubes in the magnet (0 g
*, 1 g
*, 2 g
*) and the 1 g
control tube. We found that between 20% and 50% of the genes affected by a change in effective gravity (0 g
* or 2 g
*) were present in flies in both 0 g
* and 2 g
• Approach 2: We compiled two lists of genes: one of genes that were up- or down-regulated in 0 g* compared with 1 g*, and one of genes that were altered in 2 g* compared with 1 g*.
In both approaches, we made the prior assumption that the magnetic field effects observed in the 1 g* tube (at 16.5 T) had nearly the same influence on the genes in the 0 g* and 2 g* tubes where the field was smaller (11.5 T).
The numbers of genes in the lists resulting from the two procedures described are shown in the highlighted third row of Additional file 1
, Figure S5. The results from the RPM experiment (described in the Methods section) are also shown for comparison. The five gene ontology groups with the highest statistical significance in each gene list are in the table in Additional file 1
, Figure S5. Most of the individual genes affected were not the same in the different tubes (0 g
*, 1 g
*, 2 g
*). However, we could identify a common theme in the affected gene groups
, which consisted of defence/immune/stress response and cell signalling gene ontology (GO) groups.
Figure lists those genes with an increase in expression of 2.5 fold or more, or decrease of 0.4 fold or more, at 1 g*, compared to the 1 g control (at 0 T). Those genes with the largest change in expression are listed towards the top of the table. In Figure , we list genes that have a similar significant change in expression at 0 g*, compared to 1 g*. Figure shows the same list for the change in gene expression at 2 g* compared to 1 g*. The genes listed in Figure and are those which were affected by the vertical diamagnetic force alone, and not by any other effects of the magnetic field. Since the vertical diamagnetic force altered the effective net weight of the flies, these genes may have a role in gravisensing or altered gravity adaption.
List of genes that respond to the magnetic field with a fold increase of more than 2.5 or fold decrease of more than 0.4.
There was only one gene with a similarly significant change in expression in the RPM experiment (an unknown function gene, CG15065 with a 0.365 fold expression change in the RPM). Some of the groups of genes identified in section C appear again in the lists shown in Figure . Of particular interest are genes that appear in more than one experiment (for instance, those in italics that are also present in the medium-term experiment on females, orphan arrays) owing to their additional statistical significance.
As shown in Figure , there were only two genes (of unknown function) that were significantly up-regulated at 1 g* compared to the 1 g control. Of those that were significantly down-regulated, the types of genes that appeared most frequently were those related with heat shock, immune/defence response, oxidation and lipid processes. It is interesting that the same heat shock genes that appear in this list also appeared in the 2 g* vs 1 g* list (Figure ), but in the latter they were over-expressed; the "Heat-shock-p70" gene was severely down-regulated in 1 g* compared to 1 g in males in a "medium-term" experiment, and up-regulated in 2 g* compared to 1 g* in females in a "short-term" experiment. None of the heat-shock genes that appeared in Figure appeared in the 0 g* vs 1 g* list (Figure ), but two immune response genes (Peroxiredoxin 2540 and Immune Induced molecule 23) were common to both of them, being down-regulated in 1 g* and up regulated in 0 g* experiments.
The sensitivity of these genes to the magnetic field was made more evident when we integrated these genes into a virtual cellular pathway. We used Pathway Studio 6.01 for a graphical output of the relations among the genes in each group. Additional file 1
, Figure S6 shows the magnet affected genes (Figure ), including some connector genes in order to fill the gaps amongst them. From the pathway, most of the genes were linked in less than two steps with other affected genes, suggesting that their functions were connected in the cell.
Many of the genes listed in Figure have unknown function, and it would be revealing to identify the function of each of these genes. One of these unknown-function genes deserves special attention. CG6908 appeared in both 0 g* vs. 1 g* and 2 g* vs. 1 g* lists (Figure &), and was strongly repressed (4 to 5 fold) in both conditions. Gene CG6908 encodes a relatively large protein with a PKC-like domain in the middle and was not changed due to magnetic field (Figure ). Therefore, this gene seems specifically sensitive to the magnetic field gradient (net gravity change).
We assessed how many of these magnetically-affected genes were also present in the group of 36 genes that were up- or down-regulated in the RPM long-term experiments. This result is shown in the grey box in Additional file 1
, Figure S4. We found that only one gene was up- or down-regulated in both the RPM and the "long-term" magnet experiments, compared to the relevant control (CG32641-RA; GenBank accession number). This gene encodes a protein with Heat shock protein/Chaperone DNAJ domains. Curiously, this gene was altered only at 1 g
* in the magnet, but not at 0 g
* tube, as one might expect. When we compared the group of 36 RPM-altered genes with those altered in any of the short-term experiments in the magnet, we found that three genes were commonly affected. One of these was "Yuri Gagarin", a gene selected previously as one of the gravity-response genes [38
]. The other two, automatically annotated unnamed genes, remain to be analysed further.
E) Global transcriptional states in the magnet: GEDI analysis
Taking into account that there are very few individual genes that were affected consistently between short-, medium- and long-term experiments, we analysed the transcriptome status as a whole. We analysed the microarray data with the "Gene Expression Dynamics Inspector" (GEDI) program [35
]. The GEDI software organises the gene expression patterns into mosaics of n
tiles. Each tile corresponds to a cluster of genes that behave similarly across conditions, designated a centroid. Different colours reflect the expression intensity of a centroid in each condition (in our case the average ratio of intensities compared to 1 g
controls). Additionally, GEDI places similar centroids close to each other in the mosaic, creating an image of the transcriptome and allowing its analysis as an entity by simple visual comparison of the mosaics corresponding to different conditions. For this analysis, we normalised the expression data as indicated in the supplementary methods (Additional file 1
). We used 18921 probe-sets for the GEDI analysis. They were placed in 20 × 16 mosaics with an average of 59 genes per centroid. The results obtained are available as GEDI original files in Additional file 3
and summarised in Figure .
Figure 4 GEDI 20 × 16 clustering analysis based on the three magnet experiments and one experiment at RPM. One experiment per row is shown for separate male and female data, in different Ground Based Facilities (GBF). The colour scale on the right indicates (more ...)
As expected from Results sections C and D, the transcriptome obtained from the 0 g
* tubes responded differently in each of the experiments (short-, medium- and long-term), and the response in 0 g
* (compared to 1 g
and 1 g
* controls) depended on the sex of the flies. In all cases, the response to the magnetic field was weak; most of the genes showed a log to the base 2 ratio change in expression of < 1.35 versus the control expression, as indicated by the colour scale (Figure ). We observed a similarly weak response in the other two tubes (1 g
*, 2 g
*) in the magnet. We note that the 0 g
* and 2 g
* images seem to be more closely related to each other than to the 1 g
* image. A RPM produced an effect superficially similar to that of the magnetic field (in the 1 g
* arena), in that the number of genes affected and the expression change of these genes were similar to that observed in the magnet, but the RPM affected different gene clusters (Additional file 1
, Figure S4 and S5). A change in temperature to 24°C produced a similar effect. However, the sensitivity of the transcriptome to the external physical parameter change (magnetic field, field gradient, temperature, RPM movements) depended on the biological state of the sample. For example, female flies seem to be less sensitive to short-term exposure to strong magnetic fields than males (the medium-term experiment had the opposite trend, but this could be due to the fact that the medium-term experiment with females was carried out with insufficient replicates due to time constraints). If we compare male and female patterns in the 1 g
* tube of the medium-term experiment (in Additional file 1
, Figure S4), it is possible to localise a group of genes that are up-regulated by the magnetic field in males, but down-regulated in females, suggesting differential transcriptome adaption responses to stress in selected populations of the same species.
Interestingly, the last two columns on the right of Figure , showing the 0 g
* vs. 1 g
* response and 2 g
* vs. 1 g
* response do not show opposite trends in gene expression, as one might expect initially. This is especially clear in experiments with male flies that produced very similar transcriptome images in the GEDI analyses. An opposite trend has been observed in mechanical simulators, RPM vs. centrifuge [21