Magnitude and significance of the peak of early embryonic mortality
Qinghua Chen, Zengru Di, Eduardo M. Garcia-Roger, Hui Li, Peter Richmond, Bertrand M. Roehner
aa r X i v : . [ q - b i o . T O ] M a r Magnitude and significance of the peak of early embryonic mortality
Qinghua Chen , Zengru Di , Eduardo M. Garcia-Roger , Hui Li , Peter Richmond ,Bertrand M. Roehner Version of 25 February 2020
Key-words: Embryogenesis, death rate, conception, infant death, senescence
1: School of Systems Science, Beijing Normal University, China.Email: [email protected]: School of Systems Science, Beijing Normal University, China.Email: [email protected]: Institut Cavanilles de Biodiversitat I Biologia Evolutiva, University of Val`encia, Spain.Email: [email protected]: School of Systems Science, Beijing Normal University, China.Email: [email protected]: School of Physics, Trinity College Dublin, Ireland.Email: peter [email protected]: Institute for Theoretical and High Energy Physics (LPTHE), Pierre and Marie Curie campus, Sor-bonne University, National Center for Scientific Research (CNRS), Paris, France.Email: [email protected]
Abstract
Biologically, for any organism life does not start at birth but at fertilization of theembryo. Embryonic development is of great importance because it determines con-genital anomalies and influences their severity. Whereas there is detailed qualitativeknowledge of the successive steps of embryonic development, little is known abouttheir probabilities of success or failure. Embryonic mortality as a function of postfertilization time provides a simple (albeit crude) way to identify major defects. Wefind that, in line with the few other species for which data are available, the embry-onic mortality of zebrafish has a prominent peak shortly after fertilization. This iscalled the early embryonic mortality (EEM) effect. Although a number of immediatecauses of death (e.g. infection, excess of carbon dioxide or of lactic acid, chromoso-mal defects) can be cited, the common underlying factor remains unknown.After reviewing embryonic mortality data available for chicken and a few other farmanimals, we explain that zebrafish are particularly suited for such a study becauseembryogenesis can be followed from its very beginning and can be observed easilythanks to transparent egg shells.We report the following findings. (i) The mortality peak occurs in the first 15% of the75-80 hours of embryogenesis and it is about 50 times higher than the low plateauwhich follows. (ii) The shape of the age-specific death rate is largely independent ofthe death level.Presently, little is known about the nature of embryonic defects. However, by re-viewing two special cases we show that even small initial defects, e.g. spatial cellu-lar asymmetries or irregularities in the timing of development, carry with them lethaleffects in later stages of embryogenesis.
Introduction
From defects in embryonic cell organization to lethality
The purpose of statistical physics is to derive the behavior of macroscopic systemsfrom the collective properties of its microscopic components. The present inves-tigation gives a striking illustration for it shows a massive mortality surge in earlyembryonic development which comes along with slight disturbances at cellular level.More specifically, investigations by two teams of researchers (Rideout et al. 2004,Cruz et al. 2012) have shown that anomalies in initial cell organization are earlysignals of a compromised (and eventually lethal) development. It is interesting tonote that in these two cases (to be explained more fully later on) the defects are notdue to individual cells but rather to defects in their interaction. Note that the defectconsidered in Rideout et al. is a spatial asymmetry that is fairly similar to positionaldefects in a crystallized solid.In line with some of our previous papers (Berrut et al. 2016, Richmond et al.2019a,b, Bois et al. 2019b) this study could have been entitled “The physics ofembryogenesis” because all along we keep in mind possible parallels with the work-ing and failures of physical (i.e. non living) systems; such parallels serve as guidesfor the exploration and suggest possible conjectures. We do not try to give a de-tailed description for a single species but focus on common features across speciesfor it appears that similar peaks occur in many species; in short, we study embryonicmortality as a global effect.So far, limited mortality data are available. Those recorded in our experiment for ze-brafish embryos may be the most accurate ever produced. There are two reasons forthat. The experiment was especially designed for the observation of early embryonicmortality (whereas previous experiments with farm animals had mostly an economicgoal) and it involved a sample of over 1,500 embryos. It is difficult to have samplesof this size for bigger animals like cows or pigs.The early embryonic spike is not the only mortality peak to occur in the develop-ment of an organism. In order to put it in a broader perspective, in the followingsubsections we briefly describe the other peaks.
From fertilization to death there are three phases
In previous papers a parallel was introduced between living and technical systems(Berrut et al. 2016, Bois et al. 2019a,b). From this perspective three phases can bedistinguished in human development.(1) Embryogenesis from fertilization of the embryo to birth.
Gestation time (years post fertilisation) D e a t h r a t e ( p e r ye a r a n d , e m b r y o s ) Embryo N o da t a Infant age (years post birth) D e a t h r a t e ( p e r ye a r a n d , b i r t h s ) Infant B i r t h Age (years) D e a t h r a t e ( p e r ye a r a n d , pop u l a ti o n ) Wear-out
110 1 10
10 20 30 40 50 60 70 80 90 100
Fig. 1 Evolution of the death rate in the three phases of human development.
In each phase there is a hugepeak. Whereas its existence is easy to understand in phase 2 and 3, the reason of the early embryonic mortality(EEM) peak remains a mystery. In phase 2 the decrease follows a hyperbolic power law whereas in phase 3the increase is exponential as described by Gompertz’s law (1825); in phase 1 the exact shape of the curve isnot yet known due to a lack of accuracy in empirical evidence, in particular in the first few weeks followingfertilization. The expression “wear-out phase” is borrowed from reliability engineering; it is justified by thefact that for many functions the best performance occurs between the age of 20 and 30 and then deterioratessteadily. As examples one can mention the ability to perceive sounds of high frequency or the density of bones.In most activities of common life this wear-out remains unnoticed and becomes a hindrance only in old age.Note that the maximum death rate of the senescence curve depends upon the the oldest age group selected. Asimilar observation applies to the infant curve; in this case inclusion of age groups close to birth would lift upthe maximum of the death rate. The data for phase 1 are for the island of Kauai in Hawaii in the late 1950s.The data for phase 2 are for the United States in 1960. The data for phase 3 are for the United States in 2016.The insets of panel 2 and 3 are (log,log) and (lin,log) respectively.
Sources: Embryo phase: French et al. 1962,L´eridon 1973,1977, Jarvis 2017. Infant phase: The data from zero to 12 years are presented in two differenttables (a) Under one year: Grove et al. 1968, p.210-211, (b) Over one year: Grove et al. 1968, p.318. Wear-outphase: WONDER database of the “Centers for Disease Control” (CDC), Compressed mortality, 1999-2016. (2) From birth to 10 year old there is the infant phase , also known as a wear-inphase.(3) After the age of 10 year and until senescence and death comes a phase dur-ing which the death rate increases exponentially in accordance with the well-knownGompertz’s law (Gompertz 1825, Richmond et al. 2016). In the language of relia-bility engineering such a steady increase of the failure rate is known as a wear-outprocess.Human age-specific curves for the three phases are given in Fig.1. For the sake ofbrevity in what follows the three phases will be referred to as phases 1, 2 and 3.Whereas there are accurate data for phases 2 and 3, embryonic death rates, and espe-cially the early embryonic rates are subject to a number of uncertainties as empha- The medical definition of infant mortality is mortality of newborns under one year. This definition is mostly motivatedby statistical convenience (a more detailed discussion can be found in Berrut et al. 2016, appendix A). From a biologicalperspective it is more logical to include into the infant phase the whole age interval during which the death rate decreaseswhich is 0-10 years. Note also that the ”beginning” of the death rate curve has no real signification for the followingreason. The first point is for age less than one year. However, if we would take as first point the mortality rate for age lessthan one month, or less than one day, or less than one hour, the rate would be at least 100 times higher. More details aboutthis hyperbolic behavior can be found in one of our earlier papers (Berrut et al. 2016). sized in the review paper by Gavin Jarvis (2017). The main reason is the difficultyof detection of fertilization in the first two weeks (e.g. early hormonal dosage is sub-ject to several external conditions). Similarly, early fetal deaths go unrecorded. Forinstance, in the United States only fetal deaths at 20 or more weeks of gestation arerecorded by states in the “National Vital Statistics System”. This situation suggeststwo comments • Despite the statistical uncertainties of human embryogenesis, because of itssheer magnitude (multiplication by 7 with respect to the base line), it is clear thatthere is a major mortality surge in the early phase of embryogenesis. This is calledthe early embryonic mortality (EEM) effect. • This effect can be studied in much better conditions in fish than in mammalsor birds because in the former fertilization occurs outside the body of the femaleand can be followed from the very beginning. Zebrafish eggs have the additionaladvantage that their shells are transparent, a feature that is not common to all fish,e.g. the egg shells of the killifish, another model fish used in laboratories, are not transparent.
Death rate surges in phases 2 and 3
In the embryonic and infant phase the mortality surge is at the start whereas in thesenescence phase it is at the end. Coming back to humans cases 2 and 3 are fairlyeasy to understand. • In phase 2 the newborns face many challenges such as breathing, regulatingbody temperature, being able to absorb and digest the mother’s milk. An inability toperform any one of these crucial functions means that the newborn may not live verylong. Defective lungs which were acceptable during pregnancy due to the oxygensupply provided by the mother may not be able to support the life of the newbornafter birth. Thus, birth appears to act as a filter resulting in the loss of newbornsafflicted by life threatening defects. In spite of substantial progress in neonatal as-sistance, present-day data still show a marked post-natal mortality surge. In fact,mortality during the first few days after birth did not change much over past decadesThis is because major cardiac or neurological defects can hardly be corrected.Note that, in contrast to Gompertz’s law, the post-natal fall is not an exponential buta power law (more detailed graphs can be found in Berrut et al. 2016). • The mortality surge seen in phase 3 is not a filtering effect. In engineeringlanguage it is known as a wear-out effect. According to Gompertz’s law the deathrate basically doubles every 10 years of age. With a logarithmic vertical scale thegraph is a straight line (as shown in the inset of Fig.1, panel 3); with a linear verticalscale it appears in the form of a sharp increase. Gompertz’s law implies that the sizeof a cohort decreases as an exponential whose exponent (i.e. the death rate) increases itself exponentially with age. Although it would be interesting to investigate thereasons of such a very abrupt decline, this study would bring us too far away fromour main topic and it will rather be postponed to an upcoming paper.In the following sections the paper proceeds as follows. • In addition to the human data already discussed in the introduction, we presentembryonic mortality data for the few other cases that have been studied. • We explain the design of our zebrafish experiment and we discuss its results. • In the conclusion we extend the 3-phase pattern given in the introduction. Weexplain also how embryonic mortality can be used to probe the temperature depen-dence of the successive biochemical processes of embryonic development. Thismethodology is the parallel of the injection methodology for the exploration of theDNA structure. • Finally, appendix A discusses some basic practical aspects of the experiment(e.g. the oxygen supply issue).
Insights into the embryonic mortality surge
The literature gives us almost no real clues for understanding the EEM effect. Asour only solid starting points one can recall the following facts. (i) Far from beingspecific to humans, such early spikes seem to exist for all species for which data areavailable (see below). (ii) Most often autopsies are unable to identify the immediatecauses of fetal deaths. Even in those cases where specific causes are found, they ap-pear fairly disparate (e.g. bacterial or viral infections, lack of oxygen, accumulationof ammonia or lactic acid, chromosomal anomalies) and do not reveal the commoneffect which may lie behind.Even if one accepts the fairly natural idea (borrowed from the neonatal phase) thatthe EEM is a filtering process (which in reliability engineering is called a wear-inprocess) this gives little insight. Contrary to birth defects, one cannot point to failureof specific organs for in the early embryonic phase (i.e. the cleavage phase) there areonly undifferentiated cells. That is why we think it may be enlightening to describetwo cases in which defects in early cell organization carry lethal effects.
Effect of spatial asymmetry in the cleavage phase
Fig.2 shows position anomalies which appear at the 8-cell level that is to say, as-suming a 20mn division time, some mn after fertilization. This asymmetry comestogether with unequal cell sizes.The graph shows that almost all embryos affected by such a defect will die beforehatching time. It would of course be of great interest to know the average time lagbetween the apparition of this defect and the death of the embryo. The authors of
20 40 60 80 100Abnormal embryos (%)
ASYMMETRY H a t c h i n g r a t e ( % ) Fig. 2 A position asymmetry in haddock embryos carries a lethal outcome.
Other kinds of cellular anoma-lies at the same development stage are described in the same paper. It is remarkable that cellular outcrops whereone or two cells protrude from the same cluster of cells do not have any negative effect in terms of hatching.The red segment corresponds to 1mm.
Source: Adapted from Rideout et al. (2004).
Rideout et al. did not study this aspect because the paper was written for fisheries andfor obvious economic reasons it was the hatching rate which was their main concern.In a subsequent paper we intend to study this aspect.After this example of spatial asymmetry we give an example of temporal asymmetry.
How early timing irregularities affect the viability of the embryos
If all cells replicate in a synchronous way development leads successively to 1, 2, 4,8, 16, 32, 64, . . . cells . However, observation shows deviations from synchronicity,a fact that is not surprising actually. As they have the same DNA the daughter cellsare clones of the initial zygote. However, it is known that even for clones there is adispersion in division time (Walden et al. 2016, Bois et al. 2019b). This means thatat some moments in this replication process there may be 3, 5, 6 or 7 cells (insteadof 2, 4 and 8).Will such a dispersion affect the subsequent development of the embryo? An obser-vation based on 834 human embryos shows that the answer is yes (Cruz et al. 2012).It appears that for the 552 embryos which had a successful development the timespent in a 3-cell stage was on average 0.6 hour, whereas for the 282 unsuccessfuldevelopments the time spent in a 3-cell stage was one hour. All other parametersdescribing the timing of the divisions were the same within a ± margin. In otherwords, this observation suggests that a lack of synchronicity in cell divisions had adisturbing influence on embryonic development. In fact the divisions of the zygote (i.e. the maternal ovum after its fertilization) are not the first divisions of thematernal cell. The oocyte has already the ability to divide. Under the influence of reproductive hormones a primaryoocyte completes a meiotic cell division at the end of which it splits into two separate cells: a small one which is fairlyuseless and a large secondary oocyte. Needless to say, this division can also be faulty. In this in vitro fertilization “successful” means that the embryo reached a stage (called blastocyst stage) which is justprior to its implantation in the uterus.
Here the effect is less massive than in the observation described in the previous sub-section. Clearly, it would be quite important to repeat this experiment with non-human embryos, for instance zebra embryos. Although early embryonic develop-ments in humans and fish are fairly different, one would expect a similar effect forthe simple reason that an embryo with 5 cells has necessarily a spatial asymmetry.Even if only temporary it should affect further development.
A testable conjecture for the embryonic mortality peak
Our idea of a parallel between embryogenesis and a manufacturing process mayappear somewhat speculative. However, the nice side of it is that it can be tested.Here we propose a testable conjecture. It should be noted that it is only one ofseveral possible conjectures that can be derived from such a parallel.In reliability engineering it is a common belief that, assuming identical technicalequipment, manufacturing defects increase with the speed of the production process.The reason is simple. Control procedures take time (even if they are automatic). Insuccessive biochemical reactions control mechanisms may be set up after each step.Such controls improve overall reliability but at the cost of slowing down the process.It turns out that across species there are great differences in early division times: 0.5hour for
C. elegans (a common laboratory model organism), 0.5 hour also for ze-brafish embryos, 10 hours for mice, 20 hours for cows, 24 hours for human embryos.According to the previous argument, the hatching rates of different yet ”similar”species should be in proportion of their initial division times.Clearly, the main difficulty is how to define the notion of “similar” species. For thesake of simplicity we retain the number of cells at hatching. This may be a fairlyrough criterion but at least it has the advantage of being well defined; e.g.
C. elegans has about 1,000 cells whereas zebrafish have some 25,000 cells. Thus, this argu-ment would lead us to predict a lower hatching rate for zebra fish than for
C. elegans larvae. More generally, the conjecture can be stated as follows.
Mortality conjecture.For two species having approximately the same number of cells at “birth”, earlyembryonic mortality should decrease when the initial division time increases.
Once more data become available, it will be possible to test the conjecture. If notconfirmed, it may be that our criterion based on the number of cells is too crude. Forinstance, cell diversity may also play a key-role.
Embryonic mortality in farm animals
For humans (and more generally for mammals) a distinction is made between em- bryonic development and fetal development. Because we will mainly be concernedwith zebrafish for which there is no real need for such a distinction, we will homog-enize our terminology and use “embryonic” for the whole process from fertilizationto birth.The two main sources from which data can be expected are studies of farm animals(e.g. poultry or cattle) or of organisms used in laboratories (e.g.
C. elegans or mice).
Embryonic death in chickens
The duration of embryonic development from fertilization to hatching is about 21days. Fig. 3a gives the graph of embryonic mortality based on data given in Pe ˜nuelaet al. (2018).In this study the number of deaths in each age interval was obtained in two ways.(i) Candling (that is to say screening the egg with the help of a light source). (ii)After hatching time the eggs which did not hatch were opened and the day of deathwas established. At that point the distinction between non-fertilized and fertilizedeggs could be made easily: any egg which experienced embryonic development, nomatter how small, could be considered fertilized.
Age of embryo (days post fertilization) D e a t h r a t e ( p e r da y a n d , f e r til e e gg s ) Fig. 3a Embryonic death rate in chickens.
The definition of the death rate which is used here is the same asfor infant mortality that is to say number of deaths in a given age interval divided by initial number of embryos.In this study a total of 3,240 eggs were examined and there were a total of 471 embryonic deaths which gives ahatching rate of 82%. Of these deaths, 57% occurred during the first week. Older age of the female increaseddeath rates but did not change the shape of the curve. The reasons which explain the occurrences of a maximumjust prior to hatching are explained in the text. Malposition of the embryo inside the egg was a major cause ofpre-hatching death.
Source: Pe˜nuela et al. (2018, p.6505)
Apart from the early peak there is another maximum just before hatching. It has twoexplanations. • It has been observed that one half of all chick embryos which die between day18 and 20 were in an abnormal position which did not give them access to the air cellwhich is on the blunt tip of the egg (Hutt 1929). • The lungs of chicks start to work shortly before hatching; previously the oxy-gen absorbed into the egg through the shell was transferred to the embryo through anetwork of thin blood vessels constituting a kind of rudimentary lung called the al-lantois. If the amount of oxygen delivered is insufficient, breaking the shell becomesimpossible.Because for chicks (and more generally for all avian species whose eggs have a fairlyhard shell) breaking the shell is a very challenging task, one is not surprised to see ahigh pre-hatching mortality peak. One would not expect a similar peak in zebrafishembryos which have a fairly soft egg membrane and this is indeed confirmed byobservation. However, the same post hatching (or rather post-yolk) filtering processis expected for fish and birds likewise.Similar results were reported for other avian species such as: pigeons, doves, ducks,grouse, pheasants, quails (Romanoff 1949) or turkeys (Fairchild et al. 2002, p. 262,Bois et al. 2019b).
Embryonic death in cattle
The widespread use of artificial insemination allows accurate determination of thetime of fertilization. Fig.3b shows that the mortality peak occurs shortly after fertil-ization.
Age (weeks post fertilisation) E m b r y o n i c d e a t h r a t e ( p e r w ee k a n d , ) Embryos of cows
Fig. 3b Embryonic death rate in cows.
The graph gives death rates per week of age and 1,000 successfulinseminations. The study involved 63 cows and there were 35 embryonic deaths which gives a calving rate of55%. The maximum occurred in the third week, i.e. 15 to 21 days after fertilization. The whole pregnancylasts about 280 days.
Sources: Sreenan et al. (1986), First et al. (1988). In contrast with Fig.3a, there is no late maximum. This cannot be seen on Fig.3bbecause the graph does not extend to the end of gestation which occurs around week40. However, in a paper by Wathes et al. (2016) one learns that from insemination toweek 8 the pregnancy losses are about 50% of initial pregnancies whereas the lossesbetween week 8 and the end of pregnancy are only 5%.Predominance of early embryonic mortality is also observed in other farm animals.However in this case the results are less detailed than for cows as they give only thepercentages of deaths which occur in the first third of the pregnancy. A constantdeath rate throughout embryo development would mean a share of 33% in this firstthird; instead the real percentages are as follows (First et al. 1988):mare: 87% , sow: 75%, ewe: 60%Moreover, the percentages p of total embryonic deaths as a fraction of successfulinseminations were as follows:mare: p = 40% , sow: p = 40% , ewe: p = 25% Model organisms used in laboratories
Regrettably, almost no mortality data as a function of embryonic age could be foundin the literature. The only available case concerns the European perch (Alix 2016).It reveals a huge initial peak (see Fig. 6b).It is this paucity of data which led us to explore the embryonic mortality of zebrafish.
Exploration of the mortality of zebrafish embryos
For a number of reasons zebrafish offer ideal conditions for the exploration of em-bryonic mortality.(1) As for most fish the eggs of the zebrafish are fertilized outside of the body ofthe female. Therefore, in contrast to many other organisms (e.g.
C. elegans , rotifers,birds, mammals) the embryo can be observed from the very moment of fertilizationto hatching.(2) As the eggs are transparent one can see easily what is going on inside.(3) The embryonic phase lasts about 3-4 days (depending on temperature) whichis a convenient duration. Shorter, it may not give enough time to carry out the re-quired observations. Longer, it would slow down the whole experiment without anyadditional benefit.(4) After death of the embryo the eggs become opaque and therefore can be easilyidentified(5) Finally, as a model organism actively studied in genetic research, zebrafish areraised in many university laboratories. However, we will see that despite such favorable conditions, the exploration of em-bryonic mortality raises several questions.
The question of the black eggs
In experiments involving zebrafish embryos there are usually two preliminary steps.Firstly, after breeding, the eggs are collected and then, usually some 24 hours later,the eggs which have become opaque are removed because they are dead (see Fig.4aand panel 4 of Fig.4c). This could seem to be a simple operation but in fact it raisessome questions. • At 24 hours post fertilization (hereafter, hpf) there will be three kinds of blackeggs: (i) those which were not fertilized and became black as a result, (ii) thosewhich were fertilized but were not able to develop, (iii) those which were fertilized,developed for a few hours but died at some point before becoming black too. Mostoften it is assumed that all black eggs were not fertilized. We will explain in a shortmoment how to separate (i)+(ii) from (iii). • A second question is what is the time-delay between the moment when a fertil-ized egg dies and the moment when it turns black?
Early separation of fertilized from unfertilized eggs
In order to answer the first question, one should separate the fertilized from theunfertilized embryos and do that as early as possible. A procedure is suggested bythe comparison of the pictures shown in Fig.4b (panel 3) and 4c (panel 3).
Fig. 4a Live eggs (transparent) and dead eggs (opaque) with light from above.
With the light from abovethe opaque cells appear white. On the contrary with light from below the sample, they appear black (as in thelast picture of Fig.3c).
Source: Picture taken at the “Biology Institute of Paris-Seine”.
The fertilized eggs can be easily distinguished from the non-fertilized after they havedeveloped a second hump which occurs about one hour after the eggs were producedby the female; see Fig.4b. However in the following hours the number of humps will Under a stereomicroscope with light coming from below such opaque eggs appear black, whereas with light comingfrom above they appear white (see Fig.3a). Here, we will call them “black” because this is the color which appears on thepicture. Note that at this point it cannot be strictly excluded that some fertilized eggs are nevertheless too defective to developa second hump. F, 5mnF, 60mn F, 75 hour
Fig. 4b Successive steps in the development of fertilized embryos.
A critical step which occurs about onehour after fertilization is the apparition of a second hump (picture 3) which results from the first division of theinitial embryo cell. The last picture shows the embryo shortly before hatching.
Source: Pictures (except forpanel 3) taken at the “Biology Institute of Paris-Seine”.
NF, 5mnNF, 5 hours NF, 8 hours
Fig. 4c Successive steps in the development of unfertilized embryos.
The unfertilized eggs can be obtaineddirectly from a female by a gentle massage on its belly. After that the process starts when the eggs are releasedin water. In fertilized as well as unfertilized eggs one sees the apparition of a first hump. Then, whereas thefertilized eggs develop a second hump as shown in the previous figure, in unfertilized eggs the single humpcontinues to swell while at the same time taking fairly weird shapes as shown in the third picture taken some5 hours after release in water. In a process which starts some 8 hours after release all non fertilized eggs willturn black as shown in the last picture.
Source: Pictures taken at the Institut de Biologie de Paris-Seine by AlexBois and at “Beijing Normal University” by Yi Zhang. increase to 4, 8, 16, 32 and so on. At each division the new cells become smallerbecause after division they are not given enough time for further growth. When there are 32 cells (or more) they form a large hump in which individual cells can no longerbe seen and which is not very different from the single humps of unfertilized eggs(see Fig.5, panel 2). In other words, the separation of unfertilized eggs should bestbe done between one and two hours post fertilization. As the eggs must be examinedone by one a single operator can hardly treat more than 200 eggs per hour. For largersamples the solution is to increase the number of operators or/and produce the eggsin successive batches by removing the separations in the breeding boxes with timelags of about one hour. We have been using two operators and one-hour time lagsbetween successive batches.One way to test whether the separation was done satisfactorily is to collect the dis-carded eggs and to see if all of them become black after 24 hours. In the experimentpresented in the next section our policy was to accept an egg as fertilized only if2 or 4 humps could be identified clearly. Thus, when an egg was oriented in sucha way that its humps were hidden (under it when using a stereo-microscope or orabove it when using an inverted microscope) it was rejected even though it may havebeen a fertilized egg. Such a drastic selection is necessary because it is crucial not toinclude any unfertilized egg as this would artificially inflate initial death rates. Thefact that some 95% of our discarded eggs became black after 24 hours shows that theprocedure worked fairly well. Identification of dead embryos
In following the development of the fertilized eggs the most impressive observation isthat some 6 hpf the number of new black eggs starts to increase, reaching a maximumbefore dwindling to zero around 24 hours post fertilization. Note that there may bea time lag between the moment when an embryo dies and the moment when the eggbecomes black. Our observations suggest that this time-lag increases in the courseof the eggs’s development. In the appendix we mention special cases for which thetime-lag was of the order of two days.With respect to the identification of deaths there are broadly speaking 3 phases:(1) The cleavage phase which lasts about 2 hours and during which one can countthe number of cells. We did not see any stoppage in development in this phase.(2) After that the embryo assumes a shape on which changes are not easy toidentify . This is illustrated in Fig. 5. At this stage the most frequent defect is thelack of a thin line around the ball of yolk. Observation shows that this defect leadsto death and blackening within a few hours, Say, over 400; in our experiment the largest cohort numbered 770. Of course, if one follows a single egg in the course of time even small changes may be visible but in our experimentwe follow a large number of eggs and we return to the same egg only every 5-6 hours; as in such successive observationsthe orientation of the embryo changes, small differences may not be clearly visible. Fig. 5 Transition from multi-hump to thin layer.
The multi-hump structure results from rapid cellular divi-sion. In the first picture there are only 4 cells whereas in the second the cells are so small and their number solarge that one can no longer distinguish them and count them. In the third picture the cells form a thin layeraround the yolk ball. It is in this stage (from 4 hpf to 10 hpf) that the evolution of the embryo is the mostdifficult to follow.
Source: Pictures taken at the “Department of Life Science” of “Peking University”. (3) Once the embryo assumes the shape of a fish larva the development becomeseasier to follow. The heart begins to beat some 25 hours post fertilization (hpf) andthe flow of blood becomes gradually more visible particularly in the yolk sac and inthe tail. In this phase the dead embryos can be clearly identified and counted.In other words, it is for the age interval between 3hpf and 20hpf that one has to findappropriate criteria defining death.In order to identify dead embryos prior to (and independently from) their blackeningone can use the following two criteria. • Further development is stopped. • The embryo does not move.The “no development” criterion can only be used when the development leads tochanges in shape that can be clearly identified. In this respect one can make thefollowing remarks . • As already said, in the evolution leading to 2,4,8 cells the changes in shape andin number can be clearly identified. However, in all our observations we have neverseen an egg arrested in the shape of 2, 4 or 8 cells. At first sight this may suggest thatin this age group the death rate is very low. It is certainly low but one must take intoaccount that this initial cleavage phase lasts less than two hours, thus the probabilityof seeing a death is naturally much lower than for a phase which lasts several hours. • Between 2 and 8 hpf the embryo has the shape of a thin border line around theyolk; this does hardly allow a clear identification of a stop in development. • After 10 hpf the embryo takes the shape of a larva rolled up around the ball ofyolk with a head and tail clearly visible; then, in the course of time, the tail becomesmore and more detached from the yolk. This feature can be used to identify an In order to identify a latency in development one may try to take successive pictures of each egg in its separate well.We tried this procedure on a small sample of 40 eggs. It turns out that unless it can be done automatically, this operationis quite time consuming. Moreover the pictures are not easy to use because when the embryo has moved and shows adifferent side successive pictures of the same embryo do not compare easily. arrested development.Table 1 summarizes the successive steps of the measurement procedure. Table 1: Operations for measuring the number of deaths
Time Device Operation Purpose(hpf) , . Several breeding boxes Production of eggs(some 100-150 per batch) . , . Stereomic. (x20) Examination of all eggs Separation of eggs with 1 humpfrom those with 2 (or more) . ,
96 well plate Repartition: 1 egg per well To follow the development ofthe eggs individually , h ) Light pad + Observation of all eggs Identification of black eggsx10 magnifying glass After 12h Stereomic. (x50) Identification of late comers Did development stop? After 24h Stereomic. (x50) Examination of all eggs Identification of the embryoswhose heart is not beating After 60h Light pad Identification of hatched eggsNotes: “hpf” means “hours post fertilization. Two modes of observation were alternated. The light pad surveyswere fast (for a 96-well plate, it took 5mn instead of 30mn for a microscopic observation of all the eggs), andtherefore they could be repeated often. However, microscopic examination was necessary to identify the eggswhich were on hold and to see whether or not their heart was still beating. The identification of laggards givesa first signal that something is wrong. It can lead to death or to recovery.
Experimental results for embryonic mortality rates
The goal of our experiment was to measure the mortality rate of fertilized embryos asa function of the post fertilization age of the embryos. The curves which summarizethe results are presented in Fig.6a and 6b.Three comments are in order.(1) In Fig.6a despite the differences in mortality levels the shapes of the curvesare similar. This stability is reassuring and gives credibility to the early embryonicmortality surge. This observation comes in confirmation of previous observationsmade (i) for chicken (Pe ˜nuela et al. 2018, p.6505) (ii) for turkey (Fairchild et al.2002, p.262) (iii) for European perch (Alix 2016, p.161). Based on her research atthe “Institute of Marine Research” in Bergen, Dr. Maud Alix found that for codsthere is also clear evidence for an early embryonic mortality peak (personal commu-nication, results to be published shortly).The results published for zebrafish by Yui Uchida et al. (2018) go in the same direc-tion but the effect is less spectacular due to a small sample of embryos ( n = 72 ). -1
110 0 10 20 30 40 50 60 70
Age of embryo (hours post fertilization) D e a t h r a t e ( p e r h o u r o f ag e a n d , f e r tili z e d e gg s ) Fig. 6a Embryonic mortality rates for fertilized zebrafish embryos.
Each curve is for a separate series ofobservations. The numbers of fertilized embryos were as follows. (1) curve with circles: , (2) squares: , (3) triangles: , (4) crosses: (total number= , ). Due probably to the young age of the parents(meaning that they have not yet been used for reproduction, but in fact the real reason of the fragility does notmatter) , the 4th sample involved a high percentage of “fragile” eggs which led to a huge screening process. Itis to draw attention to this difference that this curve was drawn with a dotted line. However, it is interestingto observe that despite the difference in mortality levels, the shape of the curve is the same. Note that thevertical scale is logarithmic. When necessary age intervals were merged to avoid intervals with zero deaths.All experiments were performed at 23 degree Celsius. This temperature was quite stable because it was theroom temperature in a laboratory with air conditioning and deprived of windows. The graph shows death ratescomputed with the same definition as for infant rates, i.e. number of deaths in a given age interval dividedby number of initial embryos. Source: The experiments were performed during the summer of 2019 partly atthe “Department of Life Science” of “Peking University” for the initial separation of fertilized eggs from nonfertilized eggs and partly at the “Institute of Physics” of the “Chinese Academy of Science” for subsequentobservations of the embryos. (2) One may wonder what is the main factor causing variability. The comparisonof curves 1,2,3 on the one hand and curve 4 on the other hand strongly suggeststhat the characteristics of the parents are essential. On average in our experimentssome 4 male-female pairs contributed to the production of the eggs . Such a lownumber is clearly not sufficient to achieve homogenization by mixing. In fact, whatmade sample 4 special is that it had 6 young pairs and only two “normal” pairs ofwhich only one gave eggs. Thus, there was almost no homogenization. Employingmore breeding boxes will improve homogenization and will likely eliminate the mostspurious cases. The number of breeding boxes was usually higher, say 6 or 8, but breeding did not occur in all boxes. -1
110 0 10 20 30 40 50 60 70
Age of embryo (hours post fertilization) D e a t h r a t e ( p e r h o u r o f ag e a n d , f e r tili z e d e gg s ) -1 Fig. 6b Average embryonic mortality rates for zebrafish and European perch.
The curve with the circlesis the average of the three curves 1,2,3 of Fig.6a and it corresponds to the lower horizontal scale (the wholeembryonic phase lasts 3-4 days). The error bars are ± σ which defines the confidence intervals with a probabilitylevel of 0.66. This means that if repeated 100 times under identical conditions the experiment would give resultsfalling within the error bars in 66 cases. The curve with the squares is for the European perch ( Perca fluviatilis )and corresponds to the upper horizontal scale. Note that drawing a smooth line through this sharp peak washardly possible which is why we drew straight connecting segments.
Sources: Zebrafish: same as for Fig.4a;Perca fluviatilis: Alix (2016).
Naturally, instead of achieving homogenization by mixing one can also try to obtainit by selection, for instance by using clones. However, like real twins, clones areaffected by their life path and may be identical only to some extent (more detailsabout differences in real twins are given in the appendix of Bois et al. 2019b).(3) How does the order of magnitude of the overall embryonic mortality comparewith the mortality of the larvae after hatching? The results presented in Bois et al.(2019, Fig.4a) for a period of 50 days post hatching show an average of about 0.4death per hour and 1,000 hatched larva. This is approximately the order of magnitudethat we see on the right-hand side of the curves in Fig. 6a,b. Moreover, the fact thatthe amplitude of the curves is the same for the zebrafish and for the European perchsuggests that embryonic death rates are fairly robust, remember that the perch is amuch bigger fish than the zebrafish.
Conclusion
The measurements presented in this study show that there is a huge mortality peak at the beginning of the embryonic phase. There can be two sources of mortality. (i)The initial elements (i.e. oocyte and sperm cell) may be defective. (ii) Subsequentdefects may arise either in the process of fertilization or in the early stages of thedevelopment (of which we have given two examples). From assisted to autonomous embryos
It is natural to interpret the death surge as a screening process which will ensuresuccessful subsequent development of the survivors, but why should it take place atthe beginning of the embryogenesis?Taking again inspiration from the birth process in which it is the environment changewhich triggers the screening process, we can observe that here too there is an envi-ronment change when the embryo becomes autonomous and does no longer rely onthe support of the female organism. For instance, a partially defective function mayhave been acceptable as long as it could be propped up through interaction with thefemale organism. Once released in water the fertilized embryos must rely on theirown resources. The successful completion of the second hump seems to signal thatthe embryo has passed this first screening. Fig.7 extends the graphs of Fig.1 to thepre-embryonic phase because a knowledge of this phase should give a better under-standing of the EEM.
Age D e a t h r a t e Gem cellsTotipotent cellsPluripotent cellsMultipotent cellsDifferentiated cells S t a r t o f oo cy t e g r o w t h Oocyte F e r tili z a ti o n Embryo B i r t h Infant D e a t h Wearout ? Unknown ?
Fig. 7 Evolution of the death rate in the four phases of development.
This figure illustrates a conjecturewhich generalizes Fig.1 in two ways. Firstly, it introduces an initial phase corresponding to the “manufacturing”of the spermatozoa and oocytes, respectively containing the male and female germ cells. Secondly, this processin four phases is presumed to apply to a broad spectrum of species from fish to mammals and birds. However, itexcludes organisms whose development involves metamorphosis through successive stages (e.g. larva, puppa,imago) which follow a more complicated process.
Embryonic mortality as a probe for exploring embryogenesis
We have already mentioned the EEM measurements done by a Japanese team of the University of Tokyo (Uchida et al. 2018) but we did not yet explain why this is trulya pioneering investigation. It can serve as a model for future studies in (at least) tworespects. • The results for zebrafish are paralleled by similar results for chicken and Africanfrogs. Although essential to ensure a broad range of validity of the results, suchcomparative studies are rare. The study by Uchida et al. may convince biologiststhat the benefits justify the additional work required by comparative investigations. • Although the title of the paper refers to a specific problem in evolutionary em-bryology, its results go beyond that. Indeed by submitting embryos to temporaryshocks it was found that development is more “fragile” in early than in late stagesof embryogenesis, in the sense that further development is more affected by earlythan by late shocks.Applying temporary or permanent temperature changes means that, instead of alter-ing the “design” (i.e. the DNA) one changes the conditions of “production” (i.e. theoutput of biochemical reactions). Such an analysis can be used to explore the under-lying biochemical processes which are the components of embryonic development.Note that this (shock → response) methodology is also used extensively in genetics.By observing phenotype responses to controled alterations of the DNA one can ex-plore the DNA and RNA sequences. Because such sequences can be altered in manypossible ways, these studies give an abundant harvest of data. On the other hand,due to their wide diversity, such results may be difficult to integrate into a commonframework. In contrast, via Arrhenius’s law , temperature is a common determiningfactor of biochemical reactions, and one can therefore expect that temperature effectsfollow fairly basic laws.Finally the following objective can be mentioned for upcoming experiments. Mathematical characterization of the shape of the embryonic spike
Does embryonic mortality follow an exponential or a power law decay? A reliableanswer would require more accurate death rate measurements. As the birth peakdecay is known to be hyperbolic, a similar shape for the EEM fall would give furtherconfidence in a parallel between the two phenomena. Several kinds of shocks were used: temperature changes, ultra violet light, chemical inhibitors, but in the subsequentdiscussion we focus particularly on heat shocks. It can be written: k = A exp ( − E a /RT ) k = reaction constant , T = Kelvin temperature , R = constant of ideal gases , E a = activation energyThe formula shows that the reaction constant (which controls the speed of the reaction) increases with the temperature. Atbiomolecular level, the transformations undergone by the embryo consist in biochemical reactions which is why biologicalprocesses also follow Arrhenius’s law. Appendix A. Some practical issues of the experiment
The purpose of this appendix is to explain some special aspects of the experiment inthe hope that it may be helpful for researchers who wish to repeat this experimentwith higher accuracy i.e. less background noise.
Organization of the experiment and bioethical rules
The experiment was performed in China at the Institute of Physics of the ChineseAcademy of Science in Beijing. The eggs were produced by standard breedingtechniques at the Life Science Department of Peking University. Independently, anumber of pictures were taken at the “Biology Institute of Paris-Seine”, France.As zebrafish are vertebrates, according to European regulation experimental proto-cols should get prior ethical approval. However, this rule applies only to adult fishand to larva which are more than three days old post hatching. For experimentsinvolving only embryos there are no specific rules, hence no need for prior approval.
The oxygen supply question
As explained in Table 1, after separation from the unfertilized, the fertilized eggswere distributed in a 96-well plate. Each of these wells has a diameter of 6mm and adepth of 10mm which gives a ratio r = (surface of interface)/depth=2.8mm − . It isthis ratio which rules the oxygenation process of the water (Chen et al. 2019).With respect to oxygen supply there are two distinct phases: (i) Before the heart isworking, oxygen is supplied and distributed to the embryo by diffusion. (ii) Oncethe heart works, the oxygen is still supplied by diffusion but it can be distributedmore effectively through the hemoglobin of the blood flow.Intuitively, it seems clear that the consumption of oxygen is correlated with the activ-ity of the embryo which suggests that it should become higher as the embryogenesisprogresses. This is indeed confirmed by observation (Green 2004). While embryo-genesis progresses three effects take place. • As already said the required volume of oxygen increases • Tail movements which appear on day 2 produce small displacements of the eggsand therefore convection through which the higher concentration of oxygen near thewater surface in each well will spread to the bottom of the well where the egg islocated. • After two days there is a proliferation of bacteria (clearly visible in the vicinityof the eggs) which consume oxygen.As it is difficult to draw a clear conclusion from these conflicting effects we set upa test. Instead of the 96-well plate we used a plate with 384 wells. These wells The gills start to work only some 20 days after hatching. have a 3mmx3mm square section and the same height of 10mm as the 96-well plate.This gives r = 0 . mm − , i.e. three times smaller than for the 96-well plate. Evenin such narrow wells the embryos did not die but had a small hatching rate of only30%. It can also be observed that even in the 96-plate, after hatching the larva remainmostly motionless. This may be due to insufficient oxygen but it may also be due toinsufficient space. It is true that the larva would have enough space for swimmingaround the well but at this stage their natural movements consist in straight forwardleaps for which there is not enough space.Two precautions will improve the oxygenation of the water in the wells. (1) the plateshould be left open for a cover would reduce oxygen concentration particularly inthe central part of the plate. (ii) The wells should be filled (and refilled) with wateronly to one half (or one third) of their height so as to increase the r ratio. The temperature question
For zebrafish a temperature of 28 degree Celsius is generally considered as optimum.As a consequence of the Law of Arrhenius, the speed of the process of embryogene-sis increases with temperature (at least in the interval 20-34 degrees). However, mostof our experiments were done at a constant room temperature of 23 degrees. Why? • When kept in an incubator at 28 degrees, the eggs will experience heat shocksevery time they are taken out of the incubator for observation (and similarly whenput back into it). We do not know how such recurrent temperature shifts will affectthem. • Beside the velocity effect, a temperature under (or over) 28 degrees may alsoaffect the mortality and this effect may be different across age groups. The waydifferent age groups respond may tell us something about the processes which takeplace.
Special episodes
We mention here two special episodes. We think that they should be seen as over-extended manifestations of effects which occur also in normal cases although in lessspectacular form. This is obvious for the second case and it may also be true for thefirst.
Latency effect
In three cases we have observed a latency time in the developmentprocess in the sense that during two days there was no development nor any move-ment of the embryos. Then, after this delay, in two cases the development resumednormally whereas in the third case the heart stopped and shortly after the egg became One can recall that in human embryonic development, there can be a waiting time effect in which implantation ofthe embryo in the uterus can be delayed by several days during which the embryo remains in a quiescent stage. Innatural conditions a distinction is made between diapause which is in response to adverse environmental conditions andquiescence which is rather triggered by internal factors. black. Effect of egg quality
One set of eggs was of bad quality as was apparent by thehigh initial number of eggs in bad shape (eggs of irregular shape, broken shells, eggscontaining black particles). In such a case, if the first day acts as a filter, one expectsa high early mortality. This is indeed what happened. It is true that a high mortalityfor mediocre eggs is nothing surprising; what is more remarkable is the fact that afterthe first 24h their mortality rate fell as rapidly as for normal eggs. Acknowledgments
We wish to express our sincere gratitude to the following per-sons for their help and interest: Maud Alix of the “Institute of Marine Research” inBergen, Norway; Alex Bois of “Sorbonne University”; Naoki Irie and Yui Uchidafrom the “University of Tokyo”; Xiaolong Fan, Xueri Ma and Yi Zhang from “Bei-jing Normal University”; Bo Zhang and Christopher Krueger from “Peking Univer-sity”; Chao Jiang of the “Institute of Physics” of the “Chinese Academy of Science”.
References