Microgravity

The Bion-M 1 biosatellite, launched on April 19, 2013 from cosmodrome Baikonur, and the descent module landed on May 19, 2013 in the vicinity of Orenburg, successfully fulfilling the plan for an unmanned, 30-day-long orbital spaceflight. Housing and climate parameters were replicated in the subsequent ground control (GC) experiment (July 26 to August 26, 2013). A total of 4 experimental groups were used for the flight and ground control experiments (n = 45 per group). An additional fifth group included the backup mice for the main flight group (n = 45). Mice of the space flight group (SF) were exposed to microgravity for 30 days. Concurrent with the SF mice, another group of 45 mice remained in the animal facility (SFV). The ground control (GC) experiment was conducted, after the landing, in the refurbished BOS flight habitats. The habitats were installed in a climatic chamber that replicated the temperature, humidity, gas composition and other flight-specific climate parameters. The corresponding vivarium control (GCV) mice were housed in the animal facility. The separate and concurrent GCV groups were used to account for possible seasonal differences between SF and GC mice.
Each of the groups (SF, GC, SFV, GCV and backup SF mice) included mice designated for in vivo studies and recovery (n = 10) and mice for dissection and in vitro measurements (n = 35). Each in vivo study subgroup, in its turn, consisted of 5 mice implanted with telemetry probes to monitor blood pressure and 5 intact animals.
Mice were handled and trained before the flight and ground control experiments. Basically, training consisted of shaping the groups of three mice each for social housing and adaptation to paste diet. The training of mice designated for in vivo studies was more comprehensive. It started with implanting the telemetry probes and, following recovery, a set of preliminary behavioral and functional tests (Figure 1, Table 1 and 2).

Detailed methods are available in the Online Supplementary Information. In brief, IMR90 iPSCs52 (link) and hESCs (H7 and H9)53 (link) were used to generate cardiac progenitors by growth factors2 (link)54 (link) or small molecules13 (link). Progenitor cardiac spheres were generated by forced aggregation in AggreWells28 (link) and subjected to simulated microgravity using an RPM55 (link)56 . Characterization was performed to compare cells under standard gravity and simulated microgravity using various methods28 (link)57 (link)58 (link) including microscopy, immuonocytochemical analysis, flow cytometry, qRT-PCR, RNA-seq, MEA recordings, calcium imaging, and patch-clamp.

Image analysis was performed on the pictures of larvae stained with Alcian blue for cartilage or Alizarin red for bone. Individual cartilage and bone elements were identified according to [10 (link), 15 (link), 66 (link)–68 ]. For morphometric analysis, images were uploaded into the CYTOMINE environment [69 ] and manually annotated by positioning 21 landmarks for larvae stained for cartilage (Fig 1A) as previously defined in the CYTOMINE ontology. 29 landmarks were placed for larvae stained for bone in hormonal treatments (Fig 1C), of which 15 were selected for the hypergravity experiments. The program then defines the positions of all selected landmarks and computes all the distances (in pixels) and angles (in radian) of all the possibilities between two points of interest. These data were exported into an Excel file and a selection of interesting measures was conducted by performing principal component analysis on data obtained from differently treated larvae to identify invariable or redundant measures. The measures selected were: for cartilage (Alcian blue): Anterior to Ethmoid plate, Anterior to Posterior, Articulation down to Articulation up, Ceratohyal ext. down to Ceratohyal ext. up, Ceratohyal ext. down to Ceratohyal int. down, Ceratohyal ext. up to Ceratohyal int. up, Ethmoid plate to Posterior, Hyosymplectic down to Hyosymplectic up; and for bone (Alizarin red): Anguloarticular down to Anguloarticular up, Anterior to Notochord, Anterior to Parasphenoid a, Branchiostegal ray 1 down to Branchiostegal ray 1 up, Entopterygoid down to Entopterygoid up, Maxilla down to Maxilla up, Opercle down to Opercle up, Parasphenoid a to Parasphenoid b, Parasphenoid b to Parasphenoid c, area of the parasphenoid triangle: parasphenoid a, b, and c, and finally the angles between parasphenoid a and b, a and c, b and c.
Statistics were performed using GraphPad Prism5. A t-test was used for control versus treatment experiments, while a one way ANOVA was used for multiple comparisons.
Morphometric analysis did not inform about the extent of ossification within each larva. Thus, a systematic structure analysis was generated. Each bone structure was classified based on the progress of development into one of the four following categories: absent, early ossification, advanced ossification and over ossification. When values were considered as quantitative, comparison between two groups (control versus chemical treatment or hypergravity in 1g>3g) was assessed by a Student t-test, while comparison between different treatments ("relative microgravity" experiment) was assessed by an analysis of variance (ANOVA). A contingency table considered ordinal values distributed among the 4 classes (from absent to over ossification) or only 3 classes when one class was not present in the sample. Association between classes and treatment was assessed by X² test and by an ordinal logistic regression and the odds ratio (OR). The "relative microgravity" experiment was analyzed in addition by grouping the 3g, 3g>1g and 3g>axe versus the 1g sample.
Statistical analyses were performed using the Statistica Software (version 10). Results were considered statistically significant at the 5% critical level (p < 0.05).

The AELISS is a miniaturized version of the Electrochemical Microgravity Laboratory (EML) comprised of two major components: the Electronic Rack (ER) and the Electrochemical Equipment Box (EEB)^{6 (link)–8 (link)}.
Ammonia Electrooxidation Laboratory at the ISS (AELISS) was inside a 2U Nanoracks module (NR-2U) of 4'' x 4'' x 8'' dimensions and connected to the ISS power rack via a USB-b port (5 V and 5 A). AELISS is comprised of two main parts: the electrical module and the electrochemical containers. The electrochemical containers are comprised of two sets of three-electrode electrochemical cell setup. The AP controlled the two peristaltic pumps that pumped the ammonia solution to the electrochemical flow cell and two sets of six (6) electrochemical experiments were run. AP provides a fixed potential to an electrochemical cell setup while it monitors the current as a function of time; this is a chronoamperometry experiment (CA). Afterward, on another electrochemical cell setup, the system controls a changing potential while it monitors the current generated (i.e. cyclic voltammetry (CV)). The electrochemical cell setup consisted of two sets of six screen-printed electrodes shown in Fig. 3. The working electrodes were modified with Platinum Vulcan catalyst ink^{9 (link)}. The electrochemical part is triply contained in order to avoid any unlikely leakage of diluted ammonia solution (0.05 M (NH₄)₂SO₄ in 0.18 M NaOH) following safety guidelines. This setup holds a maximum of 10 mL 0.05 M Ammonia solution at pH = 12.9. The experimental data was stored on a memory card. Figure 4 illustrates the closed Ammonia Electrooxidation Lab at the ISS (AELISS) device and Fig. 5 shows the open device picture of two sets of closed loops of ammonia solution inside the protective plastic frame and the closed device inside the Ziploc^® bag.Fig. 4

a Illustration of the closed AELISS device and b schematic Isometric view of the plastic protective frame with the electrochemical cells. The AELISS had a potentiostat and its plastic cover, plastic protector box frame, two electrochemical cell arrays, two screen-printed electrodes (SPE), two peristaltic pumps, and a storage device.

Fig. 5

a Open device picture showing the two sets of closed loops of ammonia solution inside the protective plastic frame and b the closed device inside the Ziploc® bag.

The fluid problem is governed by the Navier–Stokes equations for an incompressible flow, in which gravity is neglected, and by the energy equation. In dimensionless form, these can be written as
$$\begin{array}{rl}\mathbf{\nabla }\cdot \mathit{u}& \hspace{0.17em}=0,\end{array}$$
$$\begin{array}{rl}\frac{D\mathit{u}}{Dt}& \hspace{0.17em}=-\mathbf{\nabla }p+{\mathbf{\nabla }}^2\mathit{u},\end{array}$$
$$\begin{array}{rl}\text{and}\frac{DT}{Dt}& \hspace{0.17em}=\frac{1}{Pr}{\mathbf{\nabla }}^{2}T,\end{array}$$
where $\mathit{u}$ is the flow velocity vector, $p$ is the pressure, $T$ is the temperature and L, ν/ L, ρν²/ L², L²/ν and ΔT have been used as reference quantities for the geometrical coordinates, velocity, pressure, time (t) and temperature, respectively (L being a characteristic length). The non-dimensional temperature is defined as $(T-T_{\text{ref}})/\mathrm{\Delta }T$ , where $T_{\text{ref}}$ is a suitable reference temperature (T_cold shown in figure 1 in our case). The operator $D(\cdot )/Dt$ is the usual material derivative, while $Pr=\nu /\alpha$ is the Prandtl number, ratio between the fluid kinematic viscosity, $\nu$ , and the thermal diffusivity, $\alpha$ . This parameter is left unvaried throughout the whole study (Pr = 8, corresponding to NaNO₃). Moreover, in line with the majority of existing efforts on the study of PAS, the physical properties of the fluid are assumed to be constant. Closure of the problem, however, requires the addition of the tangential stress conditions at the interface:
$$[\mathbf{\nabla }\mathit{u}+{(\mathbf{\nabla }\mathit{u})}^{\text{T}}]\cdot \mathit{n}+\text{Re}{\mathbf{\nabla }}_{\text{s}}T=\mathbf{0},$$
where $\mathit{n}$ is the unit vector perpendicular to (pointing outward) to the interface, while ${\mathbf{\nabla }}_{\text{s}}T$ is the projection of the temperature gradient on the surface separating the liquid bridge and the surrounding environment. It should be noted that the latter is assumed to be a gas characterized by a viscosity much smaller than that of the liquid bridge.
Figure 1.

Schematic representation of the liquid bridge showing the temperature gradient and its projection along the interface (these two are coincident in this case as the interface is straight), and the unit vector perpendicular to the interface. (Online version in colour.)

Accordingly, its contribution is disregarded in the tangential stress conditions and in equations (2.1–2.3), thereby allowing the description of the multiphase interfacial flow in the framework of a single-fluid approach. Moreover, heat exchange with the gas is also neglected and the free interface is modelled as a perfect cylinder. The former assumption is generally considered valid when the two supporting discs are heated and cooled ‘symmetrically’ with respect to the ambient temperature, i.e. their temperatures are T_a+ ΔT/2 and T_a − ΔT/2, respectively, where T_a is the ambient temperature and ΔT is the overall temperature difference applied to the liquid bridge. In such circumstances, the temperature of the free surface (it is almost uniform with the exception of the changes that occur in proximity to the discs) is almost identical to that of the gas ambient, thereby minimizing the interfacial heat exchange. With regard to the latter hypothesis, the static curvature of the free interface can indeed be neglected as the liquid bridges are considered in microgravity conditions, their volume is identical to that of the corresponding cylinders having the same base and height and the wetting angle of the considered fluid is close to 90° (assuming the supporting discs to be coated with graphite, [8 (link),38 (link)]). Dynamic shape deformations are also ignored by considering that the so-called Capillary number, defined as $\text{Ca}={\sigma }_{\text{T}}\mathrm{\Delta }T/{\sigma }_{\text{o}}$ , is much smaller than 1 for the conditions considered in the present work ( ${\sigma }_{\text{o}}$ being the reference surface tension evaluated at T_cold, i.e. ${\sigma }_{\text{o}}\cong 1.15\times {10}^{-1}\text{N}{\text{m}}^{-1}$ ; ${\sigma }_{\text{T}}\cong 7\times {10}^{-5}\text{N}{\text{K}}^{-1}{\text{m}}^{-1}$ → $\text{Ca}\cong 6\times {10}^{-4}\hspace{0.17em}\mathrm{\Delta }T$ ). The dimensionless parameter $\text{Re}={\sigma }_T\mathrm{\Delta }TL/\rho {\nu }^2$ appearing in equation (2.4) is the Reynolds number based on the characteristic thermocapillary velocity $U_{\text{T}}={\sigma }_{\text{T}}\mathrm{\Delta }T/\rho \nu$ , where ${\sigma }_T=-\mathrm{\partial }\sigma /\mathrm{\partial }T$ is the derivative of the surface tension $\sigma$ with respect to the temperature, $\mathrm{\Delta }T$ is the aforementioned temperature difference between the two cylindrical rods supporting the liquid bridge, and L is the distance between them. It is usual practice to refer to the Marangoni number $\text{Ma}=\text{Re}Pr$ rather than to the Reynolds number to characterize the flow field, therefore, in the following this parameter is used instead of $\text{Re}$ to describe the results.
The additional thermal and kinematic boundary conditions for the fluid phase schematically shown in figure 1 can be turned into precise mathematical relationships as follows:
On the two supporting discs:
$$\begin{array}{rl}\hspace{0.17em}& \text{Cold disc }(y=0):\hspace{0.17em}T=0,\hspace{0.17em}\hspace{0.17em}\mathit{u}=\mathbf{0}.\end{array}$$
$$\begin{array}{rl}\hspace{0.17em}& \text{Hot disc }(y=1):T=1,\hspace{0.17em}\hspace{0.17em}\mathit{u}=\mathbf{0}.\end{array}$$
At the free surface:
$$\begin{array}{rl}\hspace{0.17em}& \mathrm{\partial }T/\mathrm{\partial }n=0\hspace{0.17em}\hspace{0.17em}(\text{adiabatic}\hspace{0.17em}\text{behaviour}).\end{array}$$
$$\begin{array}{rl}\hspace{0.17em}& \mathit{u}\cdot \mathit{n}=0\hspace{0.17em}\hspace{0.17em}\text{( no}\hspace{0.17em}\text{radial}\hspace{0.17em}\text{velocity) }.\end{array}$$

Simulations have been carried out adopting the ‘buoyantBoussinesqPimpleFoam’ transient solver available in OpenFOAM, properly complemented by the thermocapillary stress condition equation (2.4) (see e.g. [33 (link),37 (link)]). In the present case, obviously, the acceleration of gravity has been set to zero (microgravity conditions), which means that only the fluid-volume preserving abilities of the solver have been exploited (incompressible flow).
In OpenfFOAM, the governing equations for the fluid phase (i.e. equations 2.1–2.3 in our case) are discretized using the Finite Volume Method (FVM). Moreover, the solver relies on the PISO algorithm of Issa [41 (link)], where a collocated grid arrangement of the variables is used to integrate the momentum equations and enforce mass conservation. The energy equation (2.3) is subsequently solved in a segregated manner. Integration in time of both the thermo-flow field and the particles governing equation is based on the backward Euler scheme. Convective terms have been discretized using the second-order accurate, linear central-differences scheme.
All these kernels have already been used in the previous studies by Capobianchi and Lappa [32 (link),33 (link),37 (link)] to which the interested reader is referred for additional details about the numerical implementation. Here, we wish simply to recall that Capobianchi & Lappa [32 (link)] provided evidence for the reliability and accuracy of these kernels through focused comparison with the simulations by Melnikov et al. [13 (link)], and the independent numerical study by Lappa [22 (link)]. Excellent agreement was obtained in terms of fundamental properties of the supercritical Marangoni flow (the frequency of the hydrothermal wave for a liquid bridge with aspect ratio (height/diameter) A = 0.34 and Ma = 20 600) and the morphology of the emerging particle structures (for $\xi =1.85$ and St = 10⁻⁴).
As a concluding remark for this section, we wish to point out that all the numerical results presented in §3 have been obtained using a mesh having the M₁ resolution defined in the earlier study by Capobianchi & Lappa [32 (link)] for the same value of the Prandtl number considered here (i.e. Pr = 8). Such a resolution was found to provide a good compromise between computational times and accuracy over an extended range of values of the Marangoni number (see Table III in [32 (link)]). The corresponding values taken by the ratio of the maximum particle diameter over the minimum computational-cell size for the conditions examined in §3 (particle Stokes number between $5.3\times {10}^{-6}\le \mathit{S}\mathit{t}\le \text{8.5}\times {\text{10}}^{\mathit{-}\mathit{5}}$ and $6\times {10}^{-7}\le \mathit{S}\mathit{t}\le \text{3.9}\times {\text{10}}^{\mathit{-}\mathit{5}}$ for A = 0.34 and A = 0.5, respectively) are less than 1 for all the considered circumstances, the only exception being the particle with diameter 80 µm in the A = 0.34 case (for which the particle-to-cell ratio slightly exceeds the unit value, which however we still consider acceptable given the one-way nature of the particle-fluid coupling implemented here).