Native climate uniformly influences temperature-dependentgrowth rate in Drosophila embryos

Steven G. Kuntz1,∗ and Michael B. Eisen2,3

1 QB3 Institute for Quantitative Biosciences2 Department of Molecular and Cell Biology3 Howard Hughes Medical InstituteUniversity of California, Berkeley, Berkeley, CA 94720, USA* E-mail: sgkuntz@berkeley.edu

Abstract

It is well known that temperature affects both the timing and outcome of animal develop-ment, and there is considerable evidence that species have adapted so that their embryosdevelop appropriately in the climates in which they live. There have, however, been rela-tively few studies comparing development in related species with different optimal develop-mental temperatures. To determine the species-specific impact of temperature on the rate,order, and proportionality of major stages of embryonic development, we used time-lapseimaging to track the developmental progress of embryos in 11 Drosophila species at sevenprecisely maintained temperatures between 17.5◦C and 32.5◦C, and used a combination ofautomated and manual annotation to determine the timing of 34 milestones during em-bryogenesis. Developmental timing is highly temperature-dependent in all species. Tropicalspecies, including cosmopolitan species of tropical origin like D. melanogaster, acceleratedevelopment with increasing temperature up to 27.5◦C, above which growth slowing fromheat-stress becomes increasingly significant. D. mojavensis, a sub-tropical fly, exhibits anamplified slow-down with lower temperatures, while D. virilis, a temperate fly, exhibitsslower growth than tropical species at all temperatures. The alpine species D. persimilisand D. pseudoobscura grow as rapidly as tropical flies at cooler temperatures, but exhibitdiminished acceleration above 22.5◦C and have drastically slowed development by 30◦C.Though the fractional developmental time of major events is affected by heat-shock, devel-opmental stages are otherwise uniformly affected by temperature, independent of species.Our results suggest that climate has a major effect on developmental timing and compar-isons should be performed based on developmental stage rather than time.

Introduction

Drosophila species are found worldwide in most non-polar climates (Figure 1A). Of the 12species sequenced in 2007 [1] and widely used for comparative and evolutionary studies, D.melanogaster, D. ananassae, D. erecta, D. grimshawi, D. sechellia, D. simulans, D. willis-toni, and D. yakuba are all native to the tropics, though D. melanogaster, D. ananassae,

1

arX

iv:1

306.

5297

v1 [

q-bi

o.PE

] 2

2 Ju

n 20

13

and D. simulans have spread recently to become increasingly cosmopolitan [2]. D. mo-javensis is a sub-tropical species, while D. virilis is a temperate species that has becomeholarctic and D. persimilis and D. pseudoobscura are alpine species.

For tropical species, the temperature of their ecosystems remains relatively constantthroughout the year. Sub-tropical and temperate species, however, experience much largertemperature swings with much lower mean annual temperatures. Daily and microenvi-ronment fluctuations challenge fly development as well. While adult flies can easily travelbetween microenvironments and are resilient to a range of temperatures, their embryos areimmobile and cannot withstand nearly as wide a temperature range [3]. Flies may thusoptimize their behavior for a particular environment (such as laying eggs in the shade aftermidday heat in warm weather or in the sun early in the morning in cold weather). Nev-ertheless, some differences in the conditions experienced by embryos in different climatesare unavoidable.

Drosophila, like most poikilotherms, grow faster in warmer weather and slower in colderweather. Life-span roughly doubles with every 10◦C decrease in temperature [4]. However,this is only over a small window of temperature, as too high or low a temperature disruptsviability. Developmental time, often measured by egg-lay to pupation or eclosion time[5, 6, 7], exhibits similar temperature-dependent trends. Changes in temperature canaffect developmental morphology and in response, populations retain developmental trendsacross latitudinal clines [5, 8]. There also appears to be a critical period for correcting fortemperature variations [9].

Despite the significant effect of temperature on development, the precise effects ofvarying temperature on development in species adapted to different climates is very poorlycharacterized. Understanding these effects has gained importance as comparative studiesin developmental biology become more common. It is, for example, unclear how to comparespecies adapted to different climates. Should one grow each at their optimal temperaturewithout knowing whether observed differences are due to interspecies or temperature dif-ferences? Or should one grow them at the same temperature even if it is suboptimal forboth species?

To provide a rigorous context for answering such questions, and to more generally delveinto the under explored effects of temperature on development, we used precise temperaturecontrol, time-lapse imaging, and careful annotation to survey the effects of temperature onembryonic development in 11 Drosophila species from diverse climates.

2

Results

Time-lapse imaging tracks major morphological events

Time-lapse imaging of embryos from 11 Drosophila species was used to track developmentalspeed. Imaging spanned from the early embryo (pre-cellularization) to hatching. Embryoswere imaged at a constant temperature, maintained to ±0.1◦C by thermoelectric Peltierheat pumps. We varied the temperature from 17.5◦C to 32.5◦C, in 2.5◦C increments.Images were taken every one to five minutes, depending on the total time of development.A minimum of 4 embryos from each species were imaged for each temperature, for a totalof 77 conditions. In total, time-lapses were collected from over 1000 individual embryos,generating over 500GB of data.

Challenges encountered in designing the setup included permitting sufficient oxygenand humidity for the embryos. Glass was problematic due to a lack of oxygenation, so anoxygen-permeable tissue culture membrane was used instead, preventing a 28% increasein developmental time. The membrane was mounted to a copper plate to maintain thermalconduction. At higher temperatures water loss became a problem, so humidifiers were usedto increase ambient humidity.

Due to the volume of images collected, computational assistance was used to facilitateimage analysis. Images from a subset of time-lapses were manually inspected to identify34 developmental stages [10, 11] (Table 1). These images, as well as all later images, wereprocessed with a Matlab script to adjust for embryo orientation, brightness and contrast,and embryo size. These representative images were averaged to generate composite refer-ence images for each stage (Figure 2A). Correlating these composite reference images tothe images in each time-lapse computationally estimates the timing of each morphologicalstage by presenting the most morphologically similar time-point (Figure 2B). Due to fluc-tuations in ambient temperature and humidity, the focal plane through the halocarbon oilvaried significantly. Therefore, z-stacks were generated for each time-lapse and the mostin-focus plane at each time was computationally determined through image autocorrelation[12, 13]. Of the 34 events, the nine most unambiguous events were selected for refinementand further analysis (pole bud appears, membrane reaches yolk, pole cell invagination,amnioproctodeal invagination, amnioserosa exposed, clypeolabrum retracts, heart-shapedmidgut, and trachea fill) (Figure 2C,D). Using a Python-scripted graphical user interface,each of the nine events was manually verified or corrected. Timing of hatching was highlyvariable, likely due to the assay conditions following dechorionation, and used only as anindication of successful development, not as a reliable and reproducible time point.

Growth rate in D. melanogaster is correlated with temperature

Analysis of D. melanogaster across seven temperatures (Figure 3,github.com/sgkuntz/TimeLapseVideos.git video supplement) revealed a very strong de-pendence of developmental timing on temperature. From 17.5◦C to 27.5◦C, there was a

3

two-fold acceleration in developmental rate, matching the previously observed correlationof lifespan with a 10◦C change in temperature [4]. However, this trend is not uniform, asdevelopmental rate reaches a maximum at 27.5◦C. By 32.5◦C, development has begun toslow, which may be due to heat stress. By 35◦C, successful development becomes extremelyrare. Below 27.5◦C, developmental rate follows a logarithmic trajectory as the temperaturedecreases,

tdev ≈ −11.37 ln(T − 15

2.5) + 33.81

where tdev is the mean developmental time in hours post-cellularization and T is temper-ature in ◦C (R2 = 0.968).

Different species follow different growth trends in response to temperature

The remaining ten Drosophila species analyzed (Figure 4A) exhibited a range of growth rateresponses to temperature changes that grouped according to their native climates. Nonewere as resilient as D. melanogaster to high temperatures. The primary distinguishingtrait of these different responses was their correlation to other animals in a similar climaticenvironment. Though only a single sub-tropical species and single temperate species wereobserved, their trends were each distinct, though more similar to each other than to anyof the other groups, possibly reflecting a shared climate trait that is absent in alpine ortropical climates.

Like D. melanogaster, other tropical Drosophila species follow similar logarithmic growthcurves (Figure 4B, github.com/sgkuntz/TimeLapseVideos.git video supplement). They ex-hibit heat-induced slowing or stalling at 30◦C, with most species not completing develop-ment at 32.5◦C (with rare exceptions among D. simulans and D. ananassae). There existssome variation species to species, but overall they are very similar, with D. ananassae, D.simulans, and D. yakuba being among the fastest and D. erecta and D. willistoni takingslightly longer. D. sechellia and D. melanogaster track the middle of the group. The tem-perature response curves of D. ananassae and D. simulans are not statistically different(P ¿ 0.05). D. melanogaster and D. sechellia are also very similar (F = 5.056).

The sub-tropical species (D. mojavensis) and temperate species (D. virilis) followeda similar temperature dependent trend, though with a much less noticeable heat-stressresponse at 30◦C. Instead of slowing growth, they continued a roughly logarithmic re-sponse even to 30◦C, though they failed to complete development at 32.5◦C and were nobetter suited for high temperatures. The slope of temperature response for D. mojaven-sis was much steeper than what was observed in any of the other species (Figure 4C,github.com/sgkuntz/TimeLapseVideos.git video supplement). D. virilis also had a steeptemperature response (Figure 4D, github.com/sgkuntz/TimeLapseVideos.git video supple-ment), though not as steep as D. mojavensis. Overall, D. virilis took longer to develop atalmost all temperatures than any of the other species. Though D. virilis has very largeeggs, embryo shape or size did not correlate with rate of development. This was best il-

4

lustrated with D. sechellia, whose embryos are the largest by a considerable margin, butdevelop at the same rate as other tropical species. D. mojavensis, which has the smallestembryos, develops at a rate most closely resembling the much larger D. virilis.

The alpine species (D. persimilis and D. pseudoobscura) exhibit a growth rate responsedominated by extreme sensitivity to heat stress (Figure 4E,github.com/sgkuntz/TimeLapseVideos.git video supplement). They follow a much moreshallow logarithmic growth response, with growth barely accelerating from 22.5◦C to themaximum rate at 27.5◦C. At 30◦C, rather than simply leveling off their growth rate, asseen in the tropical species, there is a very steep decrease in developmental rate and a largedrop-off in embryo survival. This response to heat is more severe than any other speciesand occurs at a significantly lower temperature.

Since the temperature responses are highly reproducible, the developmental time foreach species can be predicted for future experiments (Figure 5). Between 17.5◦C and 27.5◦Cthe total developmental time can be approximated relatively accurately by a logarithmicregression (Table 2). At higher temperatures, heat stress appears to counter the logarithmictrend.

Different temperatures reveal developmental advantages for different species

Growth rates of each climate group may provide advantages in a particular setting, as thefastest or most responsive animals change. The tropical species are consistently the fastestdevelopers at all temperatures, with other species at most managing to match the tropicaldevelopmental rate. At cooler temperatures, such at 17.5◦C, the alpine species grow just asfast as the tropical species, while the temperate and sub-tropical species grow considerablyslower, taking almost 30% longer to develop (Figure 6A,github.com/sgkuntz/TimeLapseVideos.git video supplement). At 22.5◦C, which is close tocommon laboratory temperatures, the alpine species have slowed to the slower edge of thetropical species, while D. mojavensis has almost caught up with the tropical flies. D. virilisis the slowest developer at this temperature (Figure 6B,github.com/sgkuntz/TimeLapseVideos.git video supplement). At warm temperatures, suchas 27.5◦C, D. virilis is still the slowest developer, but closely approaching the alpine species.D. mojavensis has a similar developmental rate to D. erecta and is very close to the bulkof the tropical species (Figure 6C, github.com/sgkuntz/TimeLapseVideos.git video supple-ment). By 30◦C, the alpine species have slowed their development considerably and arethe slowest growing embryos. D. virilis develops at a rate close to that of the tropicalspecies and D. mojavensis is developing just as fast as the tropical species (Figure 6D,github.com/sgkuntz/TimeLapseVideos.git video supplement). The overall developmentaltrends strongly support grouping these different species by their native climate (Figure6E).

5

The fractional time of each developmental event is unaffected by non-stressful temperatures

We were particularly interested in what stages may be primarily responsible for the ex-pansion and contraction of development as the temperature shifts, as well as how this mayvary between species with different temperature responses. For instance, if morphologicalrate limits prevented a stage from accelerating, it may contribute less to temperature re-sponse. By setting the end of cellularization as the zero-point of development (and thusavoiding problems with different egg-retention times) and the filling of the trachea as theend point of development (avoiding problems with the variability of hatching), embryonicdevelopment can be normalized and the fractional developmental contribution determined.

Surprisingly, D. melanogaster exhibited no major changes in its proportional develop-mental time under any of the temperature conditions tested (Figure 7A). A very slightcontraction in the proportion of time between early development (pole bud appears) tothe end of cellularization (membrane reaches yolk) is apparent when the animals are un-der heat stress. However, this does not appear at lower temperatures. Similarly, a slightcontraction between the end of cellularization and mid-germ band retraction (amnioserosaexposure) is also apparent in some species, but is again very subtle and not statisticallysignificant.

Comparing all the different species at the same temperature revealed conserved propor-tionality across species (not undergoing significant heat stress), despite significant differ-ences in developmental rate and temperature sensitivity (Figure 7B). This can be extendedto all of the observations across all conditions (Figure 7C). This permits basic predictionsof morphological event timing based on the species and temperature (Table 3). Some minordiscrepancies exist between D. mojavensis and D. virilis, but these do not appear to besignificant.

Under heat-stress, as most clearly illustrated with the alpine species, some of the pro-portionality is disrupted (Figure 8A). This can be most clearly seen in D. ananassae, D. mo-javensis, D. persimilis, and D. pseudoobscura. The heat-stress appears to have a somewhatstochastic impact on development, as some animals take considerably longer to develop anda larger proportion of their developmental time is spent in later stages (post-germband re-traction). We did not identify any particular stage as causing this delay, but rather it ap-pears to reflect a compounding and uniform slowing of development. Early heat shock sig-nificantly disrupts development enough to noticeably affect morphology in yolk contraction,cellularization, and gastrulation (Figure 8B, github.com/sgkuntz/TimeLapseVideos.gitvideo supplement). Syncytial animals are the most sensitive to heat-shock (Figure 8C).Generating linear regressions across normalized developmental stages at different temper-atures reveals a heat-stress dependent change in the proportionality of early development(pre-cellularization) (Figure 8D). While all later stages following cellularization maintaintheir proportionality even at very high temperatures, the pre-cellularization stages takeproportionally less and less time. This indicates that at higher temperatures, some pre-

6

cellularization kinetics are affected by temperature in a manor that the animal cannotcorrect, which may lead to mortality as the temperature becomes more extreme.

Discussion

Specific embryonic temperature adaptation is not universally present asan underlying mechanism

Drosophila adaptations to native climates strongly impact their embryonic developmentalresponse to temperature. Developmental trends follow the broad categories of tropical, sub-tropical, temperate, and alpine-adapted species. These trends affect their developmentalrate and threshold for heat-tolerance, but do not differentially affect different developmentalstages.

The clustering of developmental responses in species by their native climates ratherthan their climates of collection suggests that climate adaptation arises slowly or can belost. The tested strain of D. melanogaster was collected in a temperate climate and theD. simulans strain was collected in a sub-tropical climate. Nevertheless, both performednearly identically to other tropical species and unlike native species collected nearby. Thissuggests that temperature responses are neither rapidly evolving (with D. melanogasterbeing present in the temperate United States for over 130 years [14]) nor already presentbut unactivated in tropical species.

The starkly different responses of the alpine species (both in the subgenus Sophophora,along with the tropical species) from either the sub-tropical or temperate species (bothin the subgenus Drosophila) suggests that these temperature adaptations arose indepen-dently or have been lost from some evolutionary branches. The pseudoobscura subgroup,bracketed by the tropical species, may have uniquely evolved their particular temperature-response pattern, though further study of diverse Drosophila species will be required toverify this. The slower growth at cold temperatures seen in the temperate and sub-tropicalanimals may be a specific adaptation within these flies or may have been ancestrally present,but lost in the Sophophora. Again, further studies of diverse flies from different parts ofthe phylogeny adapted to a variety of climates will be needed to determine the history ofthese adaptations.

Temperature adaptations may be protective

Because temperature-specific embryonic growth-rate adaptations appear to be unique toanimals experiencing the greatest natural temperature variation, it is likely that theseadaptations have been strongly selected and convey a protective or competitive advantagein their native ranges. This is supported by the very different responses to temperatureextremes exhibited by the non-tropical species. Their responses are likely tailored to dif-ferent conditions: alpine species must grow rapidly in the cold but rarely have to grow

7

in the heat, while the temperate and sub-tropical species deal with a very wide range oftemperatures. Tropical species, on the other hand, rarely encounter either extreme heator severe cold and require no protection from such extremes.

Therefore, alpine species may be ill-equipped for warm weather and egg-laying may beavoided in the heat. However, fast growth in cold weather may provide these species witha competitive advantage and may explain their relatively rapid growth. There may exista trade-off between cold-resistance and susceptibility to heat-stress. Sub-tropical species,on the other hand, are commonly exposed to a wide range of temperatures, includingvery warm days. Slowing growth in the cold may be a diapause-like response, allowingthe embryo to wait until the weather warms before continuing to grow. Since the earlieststages of development, especially the syncytium, are very temperature-sensitive, a mixedstrategy of selective egg-lay and temperate-responsive growth rate may provide the greatestselective advantages.

Rapid growth in all conditions, as exhibited by the tropical species, may only be advan-tageous to species used to relatively warm and consistent climates. In these cases, the fliesmay be able to discard any developmentally costly adaptations to temperature extremes.Instead, growing at the maximum possible rate at all times would likely serve the tropicalspecies best. Variations between species, with D. ananassae, D. yakuba, and D. simulansbeing marginally faster at all temperatures than D. erecta and D. willistoni may be dueto other unknown factors or minor climatic specializations.

Thermal adaptation is uniform and cell-type independent

Since temperature-controlled modification of developmental rate is not dominated or evenbiased toward a particular stage, growth must be controlled by a uniform, underlyingneuron-independent mechanism. Most well-understood temperature-sensitive regulationis controlled by neurons [15, 16], but embryonic temperature response begins long be-fore the first neurons are formed, several hours after gastrulation [17]. The uniformity oftemperature-dependent changes suggests that the mechanism affects most components ofdevelopment evenly. Therefore, it is likely due to broad, but powerful, selection on a largevariety of genes or a subset of highly influential genes. The response may be transcrip-tionally driven, as the maternal to zygotic transition (MZT) occurs in cycle 8, at the verybeginning of our analysis [18]. The change in proportionality seen in the syncytium atstressfully high temperatures may be affected by the MZT. It is possible that whatevermechanism adjusts for developmental rate is differentiated between the maternal and zy-gotic stages. The developmental rate appears to be stage-independent, as heat-shockedanimals do not continue with rapid development once the temperature is lowered.

8

Morphological stages are more reliable than clock-time for cross-speciesor cross-temperature comparisons

This study demonstrates that for comparisons between species or growth conditions, stage-based selection is more reliable, more consistent, and simpler than purely clock-time basedcomparisons. Stage-based selection can use the same criteria across different species due totheir similar morphologies, despite superficial differences in embryo size and shape. Due tothe proportionality of development that is independent of both species and temperature,stage-based comparisons will be more consistent than any other option. Clock-time com-parisons are theoretically possible, but require very precise adjustment for both the speciesand any temperature variations that occur during the experiment, as well as any adjust-ments for the temperatures at which the parents were raised. As comparative analyses ofDrosophila species become more common, the ability to precisely compare diverse speciesbecomes more and more essential.

Comparisons must also take into consideration, and generally avoid, heat-related stresson the embryos. High temperatures can induce stress-related stochasticity in developmentaltiming, which may aggravate comparisons. Also important is avoiding comparisons betweena stressed species and a non-stressed species. Higher temperatures may also lead to changesin the proportionality of the developing syncytium. While most laboratory conditions(∼20-25◦C) are amenable to all the studied species, special conditions must be dealt withcarefully. For instance, comparisons between alpine and tropical species may benefit fromthe comparisons being performed at a lower temperature, due to the rapid heat-stressresponse of the alpine species. However, due to the uniform proportionality of development,performing comparisons at two temperatures, with tropical species at a higher temperaturethan the alpine species, may have minimal drawbacks.

9

Materials and Methods

Rearing of Drosophila

Drosophila strains were reared and maintained on standard fly media at 25◦C, except forD. persimilis and D. pseudoobscura which were reared and maintained at 22◦C. To verifythat growth at lower temperatures did not affect the results, a D. melanogaster line wasraised at 18◦C for several years and its temperature response profile was observed (datanot shown). Egg-lays were performed in medium cages on 10 cm molasses plates for 1 hourat 25◦C after pre-clearing for all species except D. persimilis, which layed at 22◦C. Em-bryos were collected and dechorionated with fresh 50% bleach solution (3% hypochloritefinal) for 45 to 90 seconds (based on the species) in preparation for imaging. Dechori-onation timing was selected as the time it took for 90% of the eggs to be successfullydechorionated. This prevented excess bleaching, as many species, such as D. mojaven-sis, are more sensitive than D. melanogaster. Strains used were D. melanogaster, OreR;D. pseudoobscura, 14011-0121.94, MV2-25; D. virilis, 15010-1051.87, McAllister V46; D.yakuba, 14021-0261.01, Begun Tai18E2; D. persimilis, 14011-0111.49,(Machado) MSH3;D. simulans, 14021-0251.195, (Begun) simw501; D. erecta, 14021-0224.01, (TSC); D. mo-javensis wrigleyi, 15081-1352.22, (Reed) CI 12 IB-4 g8; D. sechellia, 14021-0248.25, (Jones)Robertson 3C; D. willistoni, 14030-0811.24, Powell Gd-H4-1; D. ananassae, 14024-0371.13,Matsuda (AABBg1).

Time-lapse Imaging

Embryos were placed on oxygen-permeable film (lumoxTMfrom Greiner Bio-one), affixedwith dried heptane glue and then covered with Halocarbon 700 oil (Sigma) [19]. The lumoxfilm was suspended on a copper plate that was temperature-regulated with two peltierplates controlled by an H-bridge temperature controller (McShane Inc., 5R7-570) with athermistor feedback, accurate to ±0.1◦C. Time-lapse imaging with bright field transmittedlight was performed on a Leica M205 FA dissecting microscope with a Leica DFC310 FXcamera using the Leica Advanced Imaging Software (LAS AF) platform. Greyscale imageswere saved from pre-cellularization to hatch. Images were saved every one to five minutes,depending on the temperature.

Event estimation

A subset of time-lapses were analyzed to obtain a series of representative images for eachof the 34 morphological events described. These images were sorted based on embryoorientation and superimposed to generate composite reference images. Images from eachtime-lapse to be analyzed were manually screened to determine the time when the mem-brane reaches the yolk, the time of trachea filling, and the orientation of the embryo. Thisinformation was fed into a Matlab script, along with the time-lapse images and the set

10

of 34 composite reference images, to estimate the time of 34 morphological events duringembryogenesis via image correlation. A correlation score was generated for each frame ofthe time-lapse. The running score was then smoothed (Savitzky-Golay smoothing filter)and the expected time window was analyzed for local maxima. Computer-aided estimateswere individually verified or corrected using a python GUI for all included data.

Statistical analysis

Statistical significance of event timing was determined by t-test with Bonferonni mul-tiple testing corrections. Median correction to remove outliers was used in determin-ing the mean and standard deviation of each developmental event. Least-squares fit-ting was used to determine the linear approximation of log-corrected developmental timefor each species. Python and Matlab scripts used in the data analysis are available atgithub.com/sgkuntz/TimeLapseCode.git.

Acknowledgements

This work was supported by a Howard Hughes Medical Institute investigator award to MBEand by NIH grant HG002779 to MBE. SGK was supported by the National Institutes ofHealth under Ruth L. Kirschstein National Research Service Award (F32-FGM101960A)from the National Institute of General Medical Sciences. We obtained flies from the Bloom-ington and UCSD stock centers. We would also like to thank T. Kaplan and P. Combs fortheir advice and assistance on data analysis and programming, Ng Wei Tian for his workon event verification, and BCK and EDKK for their support.

References

[1] Drosophila 12 Genomes Consortium, Clark A, Eisen M, Smith D, Bergman C, et al.(2007) Evolution of genes and genomes on the drosophila phylogeny. Nature 450:203-218.

[2] Markow TA, O’Grady PM (2005) Drosophila: A Guide to Species Identification andUse. Academic Press.

[3] Schnebel EM, Grossfield J (1986) Oviposition temperature range in four drosophilaspecies triads from different ecological backgrounds. American Midland Naturalist116: 25-35.

[4] Loeb J, Northrop JH (1916) Is there a temperature coefficient for the duration of life?PNAS 2: 456-7.

11

[5] James AC, Azevedo RB, Partridge L (1995) Cellular basis and developmental timingin a size cline of Drosophila melanogaster. Genetics 140: 659–666.

[6] James AC, Azevedo RB, Partridge L (1997) Genetic and environmental responsesto temperature of Drosophila melanogaster from a latitudinal cline. Genetics 146:881–890.

[7] Yamamoto A, Ohba S (1982) Strategic differences in thermal adaptation between twodrosophila species, d. virilis and d. immigrans. Oecologia 52: 333-339.

[8] Partridge L, Barrie B, Fowler K, French V (1994) Evolution and development of bodyand cell size in drosophila melanogaster in response to temperature. Evolution 48:1269-1276.

[9] Lucchetta EM, Lee JH, Fu LA, Patel NH, Ismagilov RF (2005) Dynamics of Drosophilaembryonic patterning network perturbed in space and time using microfluidics. Nature434: 1134–1138.

[10] Campos-Ortega J, Hartenstein V (1985) The embryonic development of Drosophilamelanogaster. Springer-Verlag:Berlin, New York.

[11] Foe VE, Odell GM, Edgar BA (1993) The Development of Drosophila melanogaster,Cold Spring Harbor Press, chapter 3: Mitosis and Morphogenesis in the DrosophilaEmbryo. pp. 149-300.

[12] Santos A, Ortiz de Solorzano C, Vaquero JJ, Pena JM, Malpica N, et al. (1997) Eval-uation of autofocus functions in molecular cytogenetic analysis. Journal of microscopy188: 264-72.

[13] Vollath D (1988) The influence of the scene parameters and of noise on the behaviourof automatic focusing algorithms. Journal of microscopy 151: 133-146.

[14] Keller A (2007) Drosophila melanogaster’s history as a human commensal. CurrentBiology 17: R77–R81.

[15] Lee SJ, Kenyon C (2009) Regulation of the longevity response to temperature bythermosensory neurons in caenorhabditis elegans. Current Biology 19: 715-22.

[16] Xiao R, Zhang B, Dong Y, Gong J, Xu T, et al. (2013) A genetic program promotesc. elegans longevity at cold temperatures via a thermosensitive trp channel. Cell 152:806-817.

[17] Salvaterra PM, Bournias-Vardiabasis N, Nair T, Hou G, Lieu C (1987) In vitro neu-ronal differentiation of drosophila embryo cells. Journal of Neuroscience 7: 10-22.

12

[18] Pritchard DK, Schubiger G (1996) Activation of transcription in drosophila embryosis a gradual process mediated by the nucleocytoplasmic ratio. Genes & Development10: 1131-42.

[19] Technau GM (1986) Lineage analysis of transplanted individual cell in embryos ofdrosophila melanogaster. i. the method. Roux’s Arch Dev Biol 195: 389-398.

[20] Bownes M (1975) A photographic study of development in the living embryo. JEmbryol exp Morph 33: 789-801.

13

Table 1: Major morphological events in Drosophila developmentEvent Stage [10, 20] Notes

Posterior gap appears 2 Gap between yolk and vitelline membranePole bud forms 3 Cells migrate into the posterior gapNuclei at periphery 4 Cells migrate to edgesPole cells form 4 Replication of the pole cellsYolk contraction 4 Light edge of embryo expandsCellularization begins 5 Cell cycle 14Membrane reaches the yolk 5 This is regarded as the zero time-

pointPole cells migrate 6 Pole cells begin anterior movementCephalic furrow forms 6 Dorsal and ventral furrows formPole cells invaginate 7 Pole cells enter dorsal furrowCephalic furrow reclines 8 Dorsal furrow moves posteriorlyAmnioproctodeal invagina-tion

8 Invagination approaches cephalicfold

Transversal fold formation 7 Dorsal furrows between amnioproctodeumand cephalic furrow

Anterior midgut primordial 8 Tissue thickens at anterior ventral edgeStomodeal plate forms 9 Ventral gap anterior to cephalic foldStomodeum invagination 10 Ventral furrow anterior to cephalic foldClypeolabral lobe forms 10 Dorsal, ventral furrows both presentGerm band maxima 11 Maximum extension of germbandClypeolabrum rotates 11 Clypeolabrum shifts dorsallyGnathal bud appears 12 Ventral tissue between the clypeolabrum

and cephalic folds moves anteriorlyPosterior gap 11 Gap forms before germband shorteningGermband retraction begins 12 Movement begins mid-germbandAmnioserosa exposed 12 Germband retracted to the posterior

30% of the embryoGermband retracted 13 Germband fully retractedDorsal divot 14 Dorsal gap between head and amnioserosaClypeolabrum retracts 14 Clypeolabrum pulls away from ante-

rior vitelline membraneAnal plate forms 14 Posterior depression formsMidgut unified 14 Dark circle forms at embryo’s centerClypeolabrum even with ven-tral lobes

16 Ventral lobes move anteriorly to be evenwith clypeolabrum

Gnathal lobes pinch 16 Gnathal lobes meetHead involution done 17 Head lobes complete anterior migrationHeart-shaped midgut 15 Triangular midgutConvoluted gut 16 Separation between sections of the midgutMuscle contractions 17 Head begins twitchingTrachea fills 17 Developmental end pointHatch 17 Highly variable

14

Table 2: The developmental time of embryos between 17.5◦C and 27.5◦C is a species-specific function oftemperature

Species Developmental Time* R2 95% Confidence Prediction Interval for Fu-ture Observations

D. virilis tDvir = 61.14 − 15.66 ln(T − 15) 0.990 tDvir ± 0.508√

1.35 + (ln(T − 15) − 23.26)2

D. mojavensis tDmoj = 59.45 − 16.43 ln(T − 15) 0.969 tDmoj±0.900√

1.40 + (ln(T − 15) − 21.47)2

D. willistoni tDwil = 47.81 − 12.70 ln(T − 15) 0.970 tDwil ± 0.704√

1.24 + (ln(T − 15) − 22.91)2

D. pseudoobscura tDpse = 42.23 − 9.00 ln(T − 15) 0.913 tDpse± 0.505√

1.30 + (ln(T − 15) − 22.56)2

D. persimilis tDper = 38.05 − 7.09 ln(T − 15) 0.969 tDper ± 0.288√

1.33 + (ln(T − 15) − 22.72)2

D. ananassae tDana = 45.20 − 12.54 ln(T − 15) 0.968 tDana±0.530√

1.37 + (ln(T − 15) − 21.41)2

D. yakuba tDyak = 39.91 − 9.67 ln(T − 15) 0.944 tDyak±0.638√

1.34 + (ln(T − 15) − 21.84)2

D. erecta tDere = 44.14 − 10.66 ln(T − 15) 0.956 tDere ± 0.723√

1.31 + (ln(T − 15) − 23.66)2

D. melanogaster tDmel = 44.30 − 11.40 ln(T − 15) 0.968 tDmel±0.362√

1.35 + (ln(T − 15) − 22.46)2

D. sechellia tDsec = 41.94 − 10.38 ln(T − 15) 0.963 tDsec ± 0.659√

1.28 + (ln(T − 15) − 22.74)2

D. simulans tDsim = 42.87 − 11.27 ln(T − 15) 0.956 tDsim±0.570√

1.25 + (ln(T − 15) − 22.53)2

* End of cellularization to trachea fill in hours, where T is in ◦C

Table 3: The timing of specific developmental events can be predicted as a function of total developmentaltime

Stage Event Timing (hourspost cellularization)

Percent Error

Pole bud appears tpba ≈ −0.093tdev 8%

Pole cells invaginate tpci ≈ 0.018tdev 40%

Amnioproctodeal invagination tapi ≈ 0.035tdev 18%

Amnioserosa exposed tase ≈ 0.35tdev 6%

Clypeolabrum retracts tclr ≈ 0.49tdev 4%

Heart-shaped midgut thsm ≈ 0.57tdev 12%

15

A

B

?

?

Dana

Dsec

Dwil

Dpse

Dper

Dmel

Dsim

Dere

Dmoj

Dvir

DyakDana (circumtropical)DereDmel (cosmopolitan)

DperDmoj

DpseDsecDsim (cosmopolitan)Dvir (holarctic)DwilDyak

D. virilis D. mojavensis D. willistoni

D. persimilisD. pseudoobscura

D. ananassae

D. erectaD. yakuba

D. melanogaster D. simulansD. sechellia

Tropical

Alpine

D. virilis D. mojavensis D. willistoni

D. persimilisD. pseudoobscura

D. ananassae

D. erectaD. yakubaD. melanogaster D. simulans

D. sechellia

Sub-TropicalTemperate

Expanding ranges, multiple continentsMedium to large rangeRestricted range (islands)

Figure 1: Drosophila are adapted to geographically and climatically diverse ecosystems (A)While D. melanogaster and D. simulans are cosmopolitan, D. ananassae is restricted to the tropics (green),while D. virilis is restricted from the tropics (gray). Other species are more or less found in their nativeranges, covering a variety of climates. Sites of collection are noted by arrows. (B) The phylogeny of thesequenced Drosophila species reveals the predominance of tropical species. Many of the tropical speciesare closely related, though D. willistoni is an out-group compared to the melanogaster and obscura groups.Ranges vary considerably between the species.

16

A BPosterior pole bud appears

Amnioserosa exposed

Germ band retraction

starts

Clypeolabral lobe forms

Stomodeum invagination

Stomodeal plate forms

Amniopro-cotodeal

invagination

Transversal fold

formation

Cephalic furrow

reclines

Pole cell invagination

begins

Clypeola-brum rotates

Cephalic furrow

formation

Pole cells migrate

Cell membranes reach yolk

Cellulariza-tion begins

Yolk contraction

Posterior pole cells pinch off

Nuclei at peripheri

Anterior and posterior

midgut fuse

Clypeola-brum

retracts

Midgut unified

Heart-shaped midgut

Clypeolabral and ventral lobes even

Convoluted gut

Gnathal lobes meet

Head involution

done

Muscle contractions

Trachea fills

Hatch

Figure 2: Events were predicted by computational analysis before manual verification (A) Manyimages of each stage (examples on the left) were averaged to generate composite images (lateral view on theright) for each of the developmental stages, of which 29 are shown. (B) For every time-lapse, each framewas correlated to each of the 34 composite images. (C) The running scores for 6 different events, withtheir maxima (black arrows) highlighted to reflect the estimated event time. (D) The time of amnioserosaexposure is estimated by the strong correlation at about 450 frames into the time-lapse.

17

30˚C

17.5˚C

20˚C

22.5˚C

25˚C

27.5˚C

35˚C

32.5˚C

Trachea �lls

Heart-shaped midgut

Amnioserosa exposed

Amnioproctodeal invagination

Pole cells invaginate

Membrane reaches yolk

Pole budappears

B

A Clypeolabrallobe forms

Clypeolabrumretracts

D. melanogaster

Figure 3: Developmental time of D. melanogaster varies with temperature (A) Images of de-veloping D. melanogaster embryos at each temperature are shown for a selection of stages. (B) The timeindividual animals reached the various time-points are shown, with each event being a different color. Time0 is defined as the end of cellularization, when the membrane invagination reaches the yolk.

18

E

Pole Bud Appears

Heart-shaped Midgut

Membrane Reaches the Yolk

Trachea FillAmnioproctodeal Invagination

Pole Cells Invaginate

Amnioserosa ExposedA

D E

ClypeolabralLobe Forms

ClypeolabrumRetracts

Dvir

Dwil

Dana

Dyak

Dere

Dsim

Dsec

Dmel

Dmoj

Dper

Dpse

D. mojavensisD. ananassae

D. pseudoobscuraD. virilis

pole bud appearsmembrane reaches yolk

pole cell invaginationamnioproctodeal invagination

amnioserosaclypeolabrum retracts

heart-shaped midguttrachea �ll

CB

Figure 4: Other Drosophila species have similar but non-identical temperature responses (A)Images of developing embryos of each species are shown to scale. (B) There is some variation species tospecies, but all tropical Drosophila exhibit a similar temperature response-curve to D. ananassae. (C)Sub-tropical D. mojavensis has a steeper temperature response, though a similar high temperature devel-opmental time. (D) TemperateD. virilis also has a steep response, though intermediate to the previous twogroups. (E) Alpine D. pseudoobscura has a cold response like the tropical species, but longer developmentaltimes at warmer temperatures.

19

D. erecta

D. pseudoobscuraD. persimilis

D. ananassaeD. yakuba

D. melanogasterD. sechelliaD. simulans

D. virilisD. mojavensis

D. willistoni

Figure 5: Prediction of future observations of development at different temperatures Thebehavior of developing embryos can be predicted. The mean line (green), the 50% confidence predictioninterval for future observations (solid orange line) and the 95% confidence interval (dashed orange line) areshown for each species

20

pole bud appearsmembrane reaches yolk

pole cell invaginationamnioproctodeal invagination

amnioserosaclypeolabrum retracts

heart-shaped midguttrachea �ll

A B

C D

E

Figure 6: Different temperatures favor different species (A) 17.5◦C with all species shows uni-formly long developmental times, but with D. virilis and D. mojavensis being significantly longer thanother species. (B) At 22.5◦C and (C) 27.5◦C there is considerably more variation between species. Whiledevelopmental times decrease with increasing temperature across all species, higher temperatures are un-favorable to the alpine species. (D) At 30◦C, developmental rate has stopped accelerating and the alpinespecies are seeing considerable slow-down in development time. (E) The time between the end of cellu-larization and the trachea fill are shown for all species at a range of temperatures. The climatic groups– tropical (dotted line), alpine (solid line), temperate (dashed line), and sub-tropical (dot-dashed line) –clearly stand out from one another to form four general trends.

21

B

C

A D. melanogaster

Figure 7: Proportionality of developmental stages is not affected by non-heat-stress tempera-tures (A) The developmental rate in D. melanogaster changes uniformly with temperature, not preferen-tially affecting any stage. Timing here is normalized between the end of cellularization and the filling ofthe trachea. (B) Across species, development maintains the same proportionality. D. pseudoobscura standsout as not being co-linear at higher temperatures. Instead, the later part of its development is slowed andtakes up a disproportionally long time. (C) Plotting proportionality across all species and all temperaturesreveals the approximately normally distributed proportionality of all morphological stages.

22

Survived heat-shock

Died following heat-shock

Never reached cellularization

>60

32.5˚C

27.5˚C

27.5˚C

A

C

D

Proportionality under heat stress

Lethality from 37.5˚C heatshock

B

Proportion of development

Tem

pera

ture

Num

ber o

f em

bryo

s

Figure 8: Heat-stress affects syncytial developmental proportionality and morphology (A) Atheat-stress temperatures, the proportionality of developmental stages is affected in some, but not all,embryos. (B) Heat stress in D. melanogaster at 32.5◦C affects morphology during yolk contraction andgastrulation. Embryos may exhibit asynchronous yolk-contraction (first image), uneven nuclear distributionduring cellularization (second image), or disrupted morphology during gastrulation (third image). (C)Heat shock at 37.5◦C for 30 minutes reveals embryos sensitivity prior to the completion of cellularization.Most animals that had completed cellularization survived heat-shock and continued to develop properly(blue diamonds), while no animals that had not completed cellularization prior to heat-shock survived. Allembryos that died (orange stars) exhibited severe morphological disruptions. (D) Linear regression of stagesacross different temperatures reveals that, despite significant variance in later stages (shown in coloredbars), only the pre-cellularization time point is affected by heat-stress enough to exhibit a significantlydifferent slope between higher temperatures (27.5◦C and above, yellow bar) and lower temperatures (25◦Cand below, red bar).

23