Oxygen changes drive non-uniform scaling in Drosophila melanogaster embryogenesis

Rearing and imaging of Drosophila

Drosophila melanogaster, OreR, were reared and maintained on standard fly media at 25°C. Egg-lays were performed in medium cages on 10 cm molasses plates for 1.5 hours at the temperature at which the lines were maintained after pre-clearing. Embryos were collected and dechorionated with fresh 50% bleach solution (3% hypochlorite final) for 60 seconds in preparation for imaging.

Embryos were monitored by modifying a temperature control system¹ in which an aluminum bar was embedded in an acrylic box (TAP Plastics). Both ends of the aluminum bar were external to the box and bound to Peltier heat pumps and heat sinks. A thermistor connected to the aluminum bar provided feedback to maintain the temperature using an H-bridge temperature controller (McShane Inc., 5R7-570). Embryos were glued³³ to oxygen-permeable film (lumox, Greiner Bio-one), covered with Halocarbon 700 oil (Sigma), and placed over holes drilled in the aluminum for imaging. An oxygen sensor (Grove Gas sensor (O2)) was placed in the box and connected to an external computer (Arduino-style Seeeduino V3.0 (Atmega 328P)). Finally, the box was sealed with two gas inputs and an over-pressure release. The computer utilized the oxygen sensor input and controlled two valves via NPN transistors, one connected to an oxygen tank and regulator and one connected to a nitrogen tank and regulator, to maintain specific oxygen concentrations in the box (Figure 1A).

Time-lapse imaging with bright field transmitted light was performed on a Leica M205 FA dissecting microscope with a Leica DFC310 FX camera using the Leica Advanced Imaging Software (LAS AF6000 version 2.3.5) platform. Greyscale images were saved from pre-cellularization to hatch. Z-stacked images were saved every two minutes (five minutes at 17.5°C). Analysis data available from http://dx.doi.org/10.6084/m9.figshare.1572474 and imaging data available from http://dx.doi.org/10.6084/m9.figshare.1582639.

Z-stack and image analysis were conducted as previously described¹. Events selected for measurement (pole-bud appearance, membrane reaching yolk, pole cell invagination, amnioproctodeal invagination, amnioserosa exposure, clypeolabrum retraction, clypeolabrum and ventral lobes being even, heart-shaped midgut, and the filling of the trachea) were identified by hand using a graphical user interface. Oxygen dependent trends were analyzed with least-squares regression. Significant differences between events in their response to oxygen changes were determined by comparing the pooled estimate of the variation about the regression line using a t-test with a Bonferonni multiple testing correction. For modeling total developmental response to oxygen and temperature changes, least-squares regression was used based on linear, exponential, logarithmic, polynomial (up to cubic), and inverse proportional models, with the models consistently yielding the best Pearson product-moment correlation coefficient being selected. For the combined effect of both oxygen and temperature, all possible combinations of exponential and inverse-proportional models identified for each component were attempted with least squares surface regression. The curve fit with the best adjust correlation coefficient $({\overline{R}}^2)$ across all available data was selected. All scripts are available at github.com/sgkuntz/OxygenCode.

Oxygen concentration controls developmental rate

We used automated time-lapse imaging in an airtight box with oxygen concentration control (±1%) and precise temperature control (±0.1°C) to track development using previously described methods¹. We investigated embryos raised at constant oxygen concentrations (29%, 25%, 21%, 17%, 14%, and 10% O₂) and kept at three different temperatures (17.5°C, 22.5°C, and 27.5°C), giving a total of eighteen specific conditions with over 800 embryos. A schematic of the setup is provided in Figure 1A. The actual setup is shown in Figure 1B.

In agreement with previous research, developmental rate correlates with oxygen concentrations (Figure 1C). Hyperoxia accelerates development, allowing embryos to hatch sooner than they would under normal atmospheric conditions. Hypoxia slows development in a dose-dependent fashion. As oxygen levels fall, an increasing fraction of embryos die or arrest their development. Therefore, there are fewer embryos shown in Figure 1C at lower oxygen concentrations due to low rates of successful development, despite similar numbers of animals being prepared for imaging (Table S1).

Figure 1. Developmental rate responds to oxygen concentrations.

(A) The oxygen control schematic. A thermistor embedded in the aluminum bar provides temperature data to the temperature controller, which in turn adjusts the voltage to the thermo-electric controllers (Peltier). An oxygen sensor in the airtight box provides feedback on oxygen concentrations to the gas controller, which opens and closes oxygen and nitrogen valves accordingly. Embryos are imaged in the center of the aluminum bar within the airtight box, indicated by the black dots in the schematic. (B) Image of the oxygen control setup mounted on the microscope at 20% oxygen and 17.5°C. (C) Developmental rate across all stages changes with the oxygen concentration, performed at 27.5°C. Each animal is represented with a dot, with averages represented with a large diamond. Developmental times here are zeroed on the end of cellularization.

Oxygen-dependent developmental scaling is non-uniform and temperature-dependent

By tracking and analyzing nine morphological stages as oxygen concentrations change, we identified significant differences in scaling between major morphological events. While all morphological events speed up with increasing oxygen concentrations (Figure 1C), their changes in speed are notably different. Syncytial development, as measured by the time between the appearance of the pole bud and the end of cellularization, takes proportionally less time as oxygen concentrations decrease, indicating that this stage is not slowed as much by decreasing oxygen (Figure 2A). The stages of gastrulation (end of cellularization, pole cell invagination, and amnioproctodeal invagination) are relatively uniformly affected. Germ band retraction, as measured by amnioserosa exposure, tracks subtly but inversely with syncytial development. More striking are the oxygen-dependent changes observed in head involution (clypeolabral retraction and advancing of the ventral lobe to match the clypeolabrum) and gut formation (heart-shaped midgut). While head involution takes proportionally more time as oxygen levels increase—meaning it does not slow as much as overall development in hypoxia—gut formation does the opposite. The midgut takes proportionally less time to form as oxygen levels increase, meaning it responds more strongly to increases in oxygen than overall development. This juxtaposition of behaviors leads to an inversion of when the cephalic lobes are even versus heart-shaped midgut formation. While hypoxia leads to head involution stages finishing first, hyperoxia results in the heart-shaped midgut forming first.

Surprisingly, the point of inversion varies with temperature (Figure 2). At 27.5°C, the inversion takes place at 29% oxygen, while at 17.5°C the inversion falls around 19% oxygen. This may be due to an overall shift in the oxygen response curve of heart-shaped midgut formation to proportionally later in development as temperatures fall. Supplementary Figure S1 reveals how each stage of development at each oxygen concentration changes with temperature.

Timing of death and arrest depend on both oxygen and temperature

Oxygen levels affect the stage at which embryos arrest or die. Higher concentrations of oxygen (29%) lead to more animals dying during early development, including death in the syncytium and a failure to properly gastrulate. This point of failure is similar to that observed at high temperatures with normal oxygen levels¹. Lethality at 25% oxygen is actually lower than that at 21%, which approximates atmospheric levels. Problems with development may be aggravated by the dechorionation and mounting procedure. At high temperatures (32.5°C) and high oxygen (29%), almost all embryos die very early in development (Table S1).

At lower oxygen levels there is a major shift from very early developmental arrest and death to mid-embryogenesis arrest (Figure 3). This holds true at all temperatures (especially at 10% O₂), but is most pronounced at 27.5°C, where the effects are still seen at 14% O₂. Frequently development halts during germ band retraction, preventing full exposure of the amnioserosa. The midgut primordia in these embryos routinely migrates haphazardly after arrest, coinciding with the embryo falling into morphological disarray. In embryos that pass these mid-embryogenesis stages, trachea formation often proves problematic. Commonly the trachea fails to form, which coincides with arrest late in midgut formation, following the heart-shaped midgut stage. These animals generally form functional muscle, with some twitching observed.

Figure 2. Oxygen-dependent changes vary with temperature.

Gut formation and head involution are the most strikingly oxygen concentration dependent processes, but germ band retraction and syncytial development are also affected. Both syncytial development and head involution take proportionally more time as oxygen concentrations increase. Gut formation and germ band retraction, in contrast, take proportionally less time as oxygen concentrations are raised. These trends hold true across all temperatures, though the rates of change as a function of oxygen do vary. Development is normalized here between the end of cellularization and the filling of the trachea.

Higher, but still hypoxic, oxygen levels (14% and 17%) have a significant fraction of embryos that fail to hatch. While embryonic development appears to be completed, including the filling of the trachea with air, larvae struggle to break out of their vitelline membrane yet fail to escape. While seen in all conditions, this behavior is most prevalent in these mildly hypoxic conditions.

Temperature influences oxygen’s control of developmental rate

Decreasing oxygen concentrations from 29% to 10% at any temperature lead to an additional sixteen to eighteen hours of embryogenesis (Figure 4A). This results in a different proportional change at each temperature, with nearly a 100% increase at 27.5°C and only a 50% increase at 17.5°C. This contrasts with changes in temperature, where developmental time roughly doubles over a 10°C range, regardless of the oxygen concentration (Supplementary Figure S2).

Changes in oxygen concentration have an inverse proportional effect on developmental rate. Least squares curve fitting was attempted with multiple models, including exponential models, for changes in oxygen concentration. The data most closely matched a model based on the Monod equation. The parameters of the response for embryogenesis as a whole change significantly with temperature, however the qualitative response is the same (Figure 4A):

Figure 3. The point of failure in development depends on both oxygen and temperature.

Lethality at different concentrations are shown for three different temperatures (27.5°C, 22.5°C, and 17.5°C). Lower concentrations of oxygen are more likely to exhibit failure during pre-tracheal development, with a particularly large increase in mortality between gastrulation and completion of the heart-shaped midgut (shown in red). A substantial increase in late development before trachea fill is also seen (shown in orange). Developmental arrest is frequently at germ band retraction. This is in contrast with higher oxygen concentrations, where failure is almost exclusively very early in development (shown in brick red), prior to the completion of gastrulation, or during difficulties hatching following trachea filling (shown in yellow). Highest survival is interestingly at 25% oxygen.

$$\begin{array}{l}{t}_{17.5}=\frac{280.92}{[O_2]}+30.13\\ t_{22.5}=\frac{167.39}{[O_2]}+17.58\\ t_{27.5}=\frac{204.03}{[O_2]}+8.55\end{array}$$

Fitting at each oxygen concentration (Supplementary Figure S2) yields relatively good fits using an exponential Arrhenius model. These different methods of fitting can be combined and yield the best fit as a multivariable non-additive model. The overall effect of oxygen and temperature can be combined to yield (Figure 4B):

Figure 4. Oxygen and temperature affect developmental time cooperatively.

(A) Fitting at different temperatures (27.5°C, 22.5°C, 17.5°C). The shift is very temperature-dependent. The solid red line represents a fit at each particular temperature, with 90% confidence of reproduction marked with the dashed orange line. The blue dashed line represents the fit across all temperatures, which departs from the individual temperature fits. (B) Total developmental time is affected by both temperature and oxygen levels. Each point represents an individual embryo at a given temperature and oxygen level. Color contours help visualize the transitions of increased heat (yellow to red contour) and increased oxygen (purple to blue contour).

$$t=e^{\frac{0.47[O_2]+31}{T}}(\frac{65.407}{[O_2]}+1.0)$$

Based on the fit, oxygen appears to have an effect on both the linear and exponential coefficients. This model is empirical and does not predict effective oxygen concentration as a function of temperature-dependent changes in oxygen solubility and diffusion. Increased oxygen may allow some additional growth acceleration, but the acceleration of growth rate appears to be leveling off, asymptotically approaching a maximum. At lower oxygen levels, the prevalence of arrest is expected to overtake the observed response curve.