A phylogenetically based transcriptome age index mirrors ontogenetic divergence patterns.

Abstract

Parallels between phylogeny and ontogeny have been discussed for almost two centuries, and a number of theories have been proposed to explain such patterns. Especially elusive is the phylotypic stage, a phase during development where species within a phylum are particularly similar to each other. Although this has formerly been interpreted as a recapitulation of phylogeny, it is now thought to reflect an ontogenetic progression phase, where strong constraints on developmental regulation and gene interactions exist. Several studies have shown that genes expressed during this stage evolve at a slower rate, but it has so far not been possible to derive an unequivocal molecular signature associated with this stage. Here we use a combination of phylostratigraphy and stage-specific gene expression data to generate a cumulative index that reflects the evolutionary age of the transcriptome at given ontogenetic stages. Using zebrafish ontogeny and adult development as a model, we find that the phylotypic stage does indeed express the oldest transcriptome set and that younger sets are expressed during early and late development, thus faithfully mirroring the hourglass model of morphological divergence. Reproductively active animals show the youngest transcriptome, with major differences between males and females. Notably, ageing animals express increasingly older genes. Comparisons with similar data sets from flies and nematodes show that this pattern occurs across phyla. Our results indicate that an old transcriptome marks the phylotypic phase and that phylogenetic differences at other ontogenetic stages correlate with the expression of newly evolved genes.

Full text

Page 1

LETTER
doi:10.1038/nature09632
A phylogenetically based transcriptome age index
mirrors ontogenetic divergence patterns
Tomislav Domazet-Los ˇo1,2& Diethard Tautz1
Parallels between phylogeny and ontogeny have been discussed for
almosttwocenturies,andanumberoftheorieshavebeenproposed
to explain such patterns1. Especially elusive is the phylotypic stage,
a phase during development where species within a phylum are
particularly similar to each other2–6. Although this has formerly
been interpreted as a recapitulation of phylogeny1, it is now
thought to reflect an ontogenetic progression phase2,where strong
constraints on developmental regulation and gene interactions
exist2,3.Severalstudieshaveshownthatgenesexpressedduringthis
stage evolve at a slower rate, but it has so far not been possible to
derive an unequivocal molecular signature associated with this
stage7–15. Here we use a combination of phylostratigraphy16and
stage-specific gene expression data to generate a cumulative index
that reflects the evolutionary age of the transcriptome at given
ontogenetic stages. Using zebrafish ontogeny and adult develop-
ment as a model, we find that the phylotypic stage does indeed
express the oldest transcriptome set and that younger sets are
expressed during early and late development, thus faithfully mir-
roring the hourglass model of morphological divergence2,3.
Reproductively active animals show the youngest transcriptome,
withmajordifferencesbetweenmalesandfemales.Notably,ageing
animals express increasingly older genes. Comparisons with sim-
ilar data sets from flies and nematodes show that this pattern
occurs across phyla. Our results indicate that an old transcriptome
marks the phylotypic phase and that phylogenetic differences at
other ontogenetic stages correlate with the expression of newly
evolved genes.
Theevolutionaryoriginofgenescanbetracedbysimilaritysearches
in genomes representing the whole tree of life. We have called this
approach ‘phylostratigraphy’ and have shown that meaningful com-
parisons can be derived from it16–18(see Supplementary Note 1). It is
importanttonotethattheprocedureidentifiesspecificallytheoriginof
novel genes with no traceable relation to existing genes or protein
domains (see Supplementary Note 2). Another important property
of phylostratigraphy is that it establishes a phylogenetic scale where
everygenewithinagenomehasitsphylogeneticrank.Here,usingthis
phylogenetic hierarchy, we extend this approach by linking it to all
expressed genes within the ontogenetic sequence. To link these two
hierarchies quantitatively we developed a transcriptome age index
(TAI), which integrates the age of a gene with its expression level at
a given developmental stage and sums this over all genes expressed at
the respective stage. The higher the TAI, the younger the transcrip-
tome (see Methods).
To apply the TAI for a developmental model system, we have
generated a fine-grained series of transcriptome data of zebrafish
development, covering a total of 60 stages, from unfertilized eggs to
ageing animals. Figure 1a shows the TAI profile, plotted along these
stages. The comparatively oldest transcript sets are expressed during
the late segmentation/early pharyngula stage, which is the develop-
mentalstagethatisusuallyequatedwiththephylotypicstageinzebra-
fish13.Thestartofheartpulsationsandbloodcirculationintheembryo
(24h)19is a morphological feature that approximately marks this
period of lowest TAI values. Phylogenetically younger transcriptome
sets are expressed before and after this stage. This correlates well with
theobservationthatearlyandlatestagesofchordatedevelopmentalso
show a higher morphological divergence between taxa2,3. During the
mid-larvalstageweseeasecondphasewhereoldertranscriptomesare
expressed, which corresponds to metamorphosis19. Although meta-
morphosis in fish is not as overt as in some other chordates (for
example, amphibians), it is nonetheless a phase with major changes
inmorphologyandlife-historystrategy.Itisparticularlyevidentinthe
reshaping of the fins, which change from a basal pattern that is seen
across all fish into the one that is more specific for zebrafish20. After
this stage, the transcriptome becomes younger and peaks in young
adults. Males and females show major differences in the overall age
index, with females expressing the relatively youngest genes.
Intriguingly, as animals become older, they express older genes again.
Analysing the contribution of the different phylostrata (ps) to the
general profile shows that they contribute to different extents
(Fig. 1b). Genes that have emerged before the evolution of metazoa
(ps1 to ps5) are more equally expressed throughout ontogeny, whereas
later-emerging genes contribute increasingly to the differential pattern.
Amoredetailedcontributionofthegenesfromthedifferentphylostrata
is summarized in Fig. 2. Here we have depicted the relative expression
levels for each stage for several phylostrata. This representation is only
partlycomparabletothatinFig.1b,asitdisregardstheactualnumberof
genes within a phylostratum. But this analysis allows several more spe-
cific points to be made.
Most genes that have arisen in ps1 (cellular origin) are general
enzymes and housekeeping genes, but their RNA is not highly
expressed before gastrulation (Fig. 2a), indicating that the products
of these genes are primarily stored as proteins in the egg. Intriguingly,
this is very different for the genes from ps2 to ps4, which have their
relatively highest expression levels at these early stages. This is also
indicative of a correlation between phylogenetic age of a gene and
ontogenetic use of its product.
The noticeable TAI peak during gastrulation (Fig. 1a) is mainly
generated by the genes from ps5 (evolution of metazoa, Fig. 2b).
Studies in sponges suggest that gastrulation is an embryological pro-
cess present since the onset of the metazoan evolution21, which is in
agreement with the peak of ps5 genes. In addition, we have previously
identified ps5 as the time of emergence of genes involved in cellular
interactions18, which are evidently of particular importance during
gastrulation.
Genes that have evolved during chordate evolution (ps9) are par-
ticularlyhighlyexpressedattheendofthe pharyngulastageandatthe
beginning of larval stages, before metamorphosis (Fig. 2b). This is
again a very suggestive correlation, because during this phase the
chordate body plan in zebrafish reaches, for the first time, a full func-
tional differentiation that is reflected in chordate-specific undulatory
swimming and the start of active feeding. Interestingly, ps7 genes
(evolution of bilateria) start to be strongly expressed at the beginning
1Max-Planck-Institut fu ¨r Evolutionsbiologie, August-Thienemannstrasse 2, 24306 Plo ¨n, Germany.2Laboratory of Evolutionary Genetics, Division of Molecular Biology, Rud–er Bos ˇkovic ´ Institute, Bijenic ˇka
cesta 54, P.P. 180, 10002 Zagreb, Croatia.
9 D E C E M B E R 2 0 1 0 | V O L 4 6 8 | N A T U R E | 8 1 5
Macmillan Publishers Limited. All rights reserved
©2010

Page 2

of metamorphosis, raising the possibility that metamorphosis-related
genes or processes have already originated in parallel to the formation
of bilateria. Ancient origins of hormonal signalling processes asso-
ciated with metamorphosis have indeed been proposed22.
Although comparable fine-grained data sets are currently not avail-
ableforothermodelsystems,onecanstillcomparethetrendsbasedon
available partial data sets. A good developmental transcriptome data
series exists for Drosophila, although it covers only one-third of the
expressed genes23. We have calculated the TAI for these data and find
thattheoverallpatternisindeedcomparabletozebrafish(Fig.3).Most
notably, the relatively oldest transcriptome is expressed during germ-
band elongation, which can be equated to the phylotypic phase in
arthropods24. Thus, this molecular signature is qualitatively compar-
able to the zebrafish data, but there are more novel genes among the
post-embryonically expressed genes in Drosophila than in zebrafish,
reflected in larger TAI values from differentiation stages onwards
(Fig. 3a). Again, we see a major difference between males and females
0
0.5
1
1.5
2
2.5
3
Unf_egg . . . . . . . . . . . . . . . . . . 1 yr 9 m
Ontogeny
Transcriptome age index
ps14
ps13
ps12
ps11
ps10
ps9
ps8
ps7
ps6
ps5
ps4
ps3
ps2
ps1
2.05
2.15
2.25
2.35
2.45
2.55
2.65
2.75
Ontogeny
Embryo
Larva
Juvenile
Adult
a
b
Euteleostomi (Osteichthyes)
Craniata (Vertebrata)
Chordata
Deuterostomia
Bilateria
Eumetazoa
Metazoa
5
(436)
Opisthokonta
Eukaryota
Cell. org.
Olfactores (Craniata + Urochordata)
Actinopterygii
Danio
1
2
3
4
6
7
8
9
10
(53)
11
(309)
12
(422)
13
(126)
14
(158)
c
(4,775)
(442)
(277)
(1,618)
(536)
(24)
(121)
(6,891)
Old
Young
Holozoa (Metazoa+allies)
Transcriptome age index
Zygote
Cleavage
Blastula
Gastrula
Segmentation
Pharyngula
Hatching
Larva
Juvenile
Adult
Male
Female
Unf_egg
45 min
1 h 45 min2 h 45 min
4 h
5 h 20 min
7 h9 h
10 h 20 min
11 h 40 min
13 h 15 h17 h 19 h
21 h
23 h
1 d 3 h
1 d 10 h 1 d 18 h
2 d 12 h
4 d8 d
14 d24 d40 d 55 d80 d
3 m 15 d
7 m
1 yr 2 m1 yr 9 m
Figure 1 | Transcriptome age profiles for the zebrafish ontogeny.
a, Cumulative transcriptome age index (TAI) for the different developmental
stages. The pink shaded area represents the presumptive phylotypic phase in
vertebrates. The overall pattern is significant by repeated measures ANOVA
(P52.4310215, after Greenhouse–Geisser correction P50.024). Grey
shaded areas represent6the standard error of TAI estimated by bootstrap
analysis. b, Transcriptome indices split according to the origin of the genes
from the different phylostrata, based on the same developmental series as in
a. c, Depiction of the phylostrata analysed; numbers in parentheses denote the
number of array probes analysed for each phylostratum.
0.0
0.5
1.0
ClBlG
ZSePhHLarvaJuvAdult
a
Relative expression level
0.0
0.5
1.0
b
ClBlGZ Se PhHLarva JuvAdult
(5) Metazoa
(6) Eumetazoa
(7) Bilateria
(9) Chordata
(1) Cellular organisms
(2) Eukaryota
(3) Opisthokonta
(4) Holozoa
Relative expression level
Figure 2 | Relative expression of the genes from each phylostratum across
the zebrafish ontogeny (same stages as in Fig. 1) for selected phylostrata
withsignificantdifferences.SeeSupplementaryFig.3 forrepresentationofall
phylostrata and significance assessments. For easier comparisons, the relative
expression calculated in relation to the highest (0) and lowest (1) expression
values across developmental stages is shown (see Methods). Bl, blastula; Cl,
cleavage; G, gastrula; H, hatching; Juv, juvenile; Ph, pharyngula; Se,
segmentation; Z, zygote.
RESEARCH LETTER
8 1 6 | N A T U R E | V O L 4 6 8 | 9 D E C E M B E R 2 0 1 0
Macmillan Publishers Limited. All rights reserved
©2010

Page 3

after hatching, but in contrast to zebrafish, the males express the
younger transcriptomes. Similar to the situation in zebrafish, ageing
Drosophila express increasingly older transcriptomes (Fig. 3a).
Breaking this pattern down to the contribution from the different
phylostrata shows again that the oldest genes contribute little to the
differential pattern, whereas genes that have emerged in ps9 (equival-
ent to the evolution of Arthropods) and later add increasingly to the
final profile (Fig. 3b).
Comparable,butevenmorelimited,ontogenetictranscriptomedata
are also available for the nematode Caenorhabditis elegans25and the
mosquito Anopheles26. The same trends can be seen for those as well,
namely the oldest genes expressed during the embryonic stages, the
youngest towards adult stages and older genes in ageing animals
(Supplementary Figs 1 and 2).
Theseconsistentoverallpatternsacrossphyla,aswellasthedetailed
analysis within zebrafish, suggest that there is a link between evolu-
tionary innovations and the emergence of novel genes16,27,28.
Adaptations are expected to occur primarily in response to altered
ecological conditions. Juvenile and adults interact much more with
ecological factors than embryos, which may even be a cause for fast
postzygoticisolation29.Similarly,thezygotemayalsoreacttoenviron-
mentalconstraints,forexample,viatheamountofyolkprovidedinthe
egg. In contrast, mid-embryonic stages around the phylotypic phase
arenormallynotindirectcontactwiththeenvironmentandarethere-
foreless likelyto be subject to ecological adaptationsand evolutionary
change. As already suggested by Darwin (discussed in ref. 15), this
alone could explain the lowered morphological divergence of early
ontogenetic stages compared to adults, which would obviate the need
to invoke particular constraints. Alternatively, the constraint hypo-
thesis would suggest that it is difficult for newly evolved genes to
become recruited to strongly connected regulatory networks12,13,15.
Thefact thatageinganimalsreverttooldertranscriptomesisinline
with the notion that animals beyond the reproductive age are not
‘visible’tonaturalselectionandcanthereforenotbesubjecttospecific
adaptationsanymore.Also,thefact that majorTAIdifferencescanbe
seen between males and females could have been anticipated, because
sexual selection is expected to continuously change phenotypic traits
between them. However, the fact that the differences go in opposite
directionsinzebrafishandDrosophilaissurprising.Wehavetherefore
studiedindetailwhichphylostratacontributemosttothesedifferences
(Fig. 4). Both taxa show a female expression bias of ps2 genes, which
maybecorrelatedtoeggproduction,asRNAfromsuchgenesisstored
in the eggs (see above). But they strongly deviate at other phylostrata.
Zebrafish shows a strong female bias of ps6 and ps12 genes, which is
absent in Drosophila (Fig. 4). Drosophila, on the other hand, shows an
extreme bias of ps14 genes in males (Fig. 4), which is caused by the
manyorphangenesinvolvedinspermatogenesis30.Thus,incontrastto
theontogeneticsimilaritiesoftheTAItrendsbetweenthetwotaxa,the
sexdifferencesareratherincongruentandindicatedifferentevolution-
ary trajectories for male–female differences.
Our study provides strong molecular support for a correlate
between phylogeny and ontogeny, as well as the hourglass model of
(1,614)
(1,034)
(83)
(65)
(49)
(170)
(135)
(24)
(20)
(37)
(42)
(38)
(45)
(194)
2.8
3
3.2
3.4
3.6
3.8
4
4.2
Ontogeny
Larvae
Pupae
Adult
Embryo
Cleavage
Blastoderm
Gastrulation
Germ band elongation
Germ band retraction
Head involution
Differentiation
Larvae
Pupae
Adult
Adult female
Adult male
0
0.5
1
1.5
2
2.5
3
3.5
4
Unf_egg . . . . . . . . . . . . . . . . . A30 d
Ontogeny
Transcriptome age index
ps14
ps13
ps12
ps11
ps10
ps9
ps8
ps7
ps6
ps5
ps4
ps3
ps2
ps1
Protostomia
Bilateria
Arthropoda
Insecta
Pancrustacea
Diptera
Drosophila
Endopterygota
Eumetazoa
Metazoa
Opisthokonta
Eukaryota
Cell. org.
1
Holozoa (Metazoa+allies)
2
3
4
5
6
7
8
9
10
11
12
13
14
a
bc
Transcriptome age index
Old
Young
Unf_egg
1 h2 h3 h 4 h5 h 6 h 7 h9 h
11 h13 h 15 h17 h19 h 21 h23 h
L33 h L49 hL67 hL84 h
L105 h
M2 hM6 h
M10 hM16 hM24 hM36 hM48 hM72 h M96 h
A3 d
A10 dA20 dA30 d
Figure 3 | Transcriptome age profiles for the Drosophila ontogeny, based
on the data in ref. 23. a, Cumulative transcriptome index for the different
developmental stages. The pink shaded area represents the presumptive
phylotypic phase in insects. The overall pattern of differences in TAI is
significantbyrepeatedmeasuresANOVA(P52.5310293,afterGreenhouse–
Geisser correction P51.22 3 10211). Grey shaded areas represent6the
standarderrorofTAIestimatedbybootstrapanalysis.b,cSameasforFig.1b,c.
LETTER RESEARCH
9 D E C E M B E R 2 0 1 0 | V O L 4 6 8 | N A T U R E | 8 1 7
Macmillan Publishers Limited. All rights reserved
©2010

Page 4

development.Under thisscheme,the phylotypicphase canbe defined
as the ontogenetic progression during which the oldest gene set is
expressed, either because this is the phase with the lowest opportunity
for lineage-specific adaptations, or because it is internally so con-
strained that newly evolved genes cannot become integrated.
METHODS SUMMARY
The TAI is the weighted mean of phylogenetic ranks (phylostrata) and is calcu-
lated for every ontogenetic stage s as follows:
TAIS~
P
i~1ei
n
i~1psiei
P
n
wherepsiisanintegerthatrepresentsthephylostratumofthegenei(forexample,
1,theoldest;14,theyoungest),eiisthemicroarraysignalintensityvalue(obtained
from Agilent Zebrafish (V2) Gene Expression Microarrays) of the gene i that acts
asweightfactorandnisthetotalnumberofgenesanalysed.Thiswayofcalculating
the index gives an increasingly stronger weight to younger phylostrata, thus com-
pensatingforthefactthattheolderphylostratausuallyharbourthelargernumber
of genes16–18.
Full Methods and any associated references are available in the online version of
the paper at www.nature.com/nature.
Received 2 September; accepted 27 October 2010.
1.
2.
Gould, S. J. Ontogeny and Phylogeny (Harvard Univ. Press, 1977).
Duboule, D. Temporal colinearity and the phylotypic progression: a basis for the
stability of a vertebrate Bauplan and the evolution of morphologies through
heterochrony. Dev. Suppl.135–142 (1994).
Raff, R. A. The Shape of Life: Genes Development, and the Evolution of Animal Form
(Univ. Chicago Press, 1996).
Richardson, M. K. et al. There is no highly conserved embryonic stage in the
vertebrates: implicationsfor current theories ofevolution and development. Anat.
Embryol. (Berl.) 196, 91–106 (1997).
Hall, B. K. Phylotypic stage or phantom: is there a highly conserved embryonic
stage in vertebrates? Trends Ecol. Evol. 12, 461–463 (1997).
Bininda-Emonds, O. R., Jeffery, J. E. & Richardson, M. K. Inverting the hourglass:
quantitative evidence against the phylotypic stage in vertebrate development.
Proc. R. Soc. Lond. B 270, 341–346 (2003).
Hanada, K., Shiu, S. H. & Li, W. H. The Nonsynonymous/synonymous substitution
rateratioversustheradical/conservativereplacementrateratiointheevolutionof
mammalian genes. Mol. Biol. Evol. 24, 2235–2241 (2007).
3.
4.
5.
6.
7.
8.Comte, A., Roux, J. & Robinson-Rechavi, M. Molecular signaling in zebrafish
developmentandthevertebratephylotypicperiod.Evol.Dev.12,144–156(2010).
Yassin, A., Lienau, E. K., Narechania, A. & DeSalle, R. Catching the phylogenic
history through the ontogenic hourglass: a phylogenomic analysis of Drosophila
body segmentation genes. Evol. Dev. 12, 288–295 (2010).
10. Davis, J. C., Brandman, O. & Petrov, D. A. Protein evolution in the context of
Drosophila development. J. Mol. Evol. 60, 774–785 (2005).
11. Hazkani-Covo,E.,Wool,D.& Graur,D.Insearchofthevertebratephylotypicstage:
a molecular examination of the developmental hourglass model and von Baer’s
third law. J. Exp. Zoolog. B 304, 150–158 (2005).
12. Irie, N. & Sehara-Fujisawa, A. The vertebrate phylotypic stage and an early
bilaterian-related stage in mouse embryogenesis defined by genomic
information. BMC Biol. 5, 1 (2007).
13. Roux,J.&Robinson-Rechavi,M.Developmentalconstraintsonvertebrategenome
evolution. PLoS Genet. 4, e1000311 (2008).
14. Cruickshank, T. & Wade, M. J. Microevolutionary support for a developmental
hourglass: gene expression patterns shape sequence variation and divergence in
Drosophila. Evol. Dev. 10, 583–590 (2008).
15. Artieri, C. G., Haerty, W. & Singh, R. S. Ontogeny and phylogeny: molecular
signaturesofselection,constraint,and temporalpleiotropy inthe developmentof
Drosophila. BMC Biol. 7, 42 (2009).
16. Domazet-Los ˇo, T., Brajkovic ´, J. & Tautz, D. A phylostratigraphy approach to
uncover the genomic history of major adaptations in metazoan lineages. Trends
Genet. 23, 533–539 (2007).
17. Domazet-Los ˇo, T. & Tautz, D. An ancient evolutionary origin of genes associated
with human genetic diseases. Mol. Biol. Evol. 25, 2699–2707 (2008).
18. Domazet-Los ˇo,T.&Tautz,D.Phylostratigraphictrackingofcancergenessuggests
a link to the emergence of multicellularity in metazoa. BMC Biol. 8, 66 (2010).
19. Parichy, D. M., Elizondo, M. R., Mills, M. G., Gordon, T. N. & Engeszer, R. E. Normal
table of postembryonic zebrafish development: staging by externally visible
anatomy of the living fish. Dev. Dyn. 238, 2975–3015 (2009).
20. Grandel, H. & Schulte-Merker, S. The development of the paired fins in the
zebrafish (Danio rerio). Mech. Dev. 79, 99–120 (1998).
21. Leys, S. P. & Eerkes-Medrano, D. Gastrulation in calcareous sponges: In search of
Haeckel’s Gastraea. Integr. Comp. Biol. 45, 342–351 (2005).
22. Flatt,T.,Moroz,L.L.,Tatar,M.&Heyland,A.Comparingthyroidandinsecthormone
signaling. Integr. Comp. Biol. 46, 777–794 (2006).
23. Arbeitman, M. N. et al. Gene expression during the life cycle of Drosophila
melanogaster. Science 297, 2270–2275 (2002).
24. Sander, K. In Development and Evolution: the sixth Symposium of the British Society
for Developmental Biology (eds Goodwin, B. C., Holder, N. & Wylie, C. C.) 137–159
(Cambridge Univ. Press, 1983).
25. Hill, A. A., Hunter, C. P., Tsung, B. T., Tucker-Kellogg, G. & Brown, E. L. Genomic
analysis of gene expression in C. elegans. Science 290, 809–812 (2000).
26. Koutsos, A. C. et al. Life cycle transcriptome of the malaria mosquito Anopheles
gambiaeandcomparisonwiththefruitflyDrosophilamelanogaster.Proc.NatlAcad.
Sci. USA 104, 11304–11309 (2007).
27. Domazet-Los ˇo, T. & Tautz, D. An evolutionary analysis of orphan genes in
Drosophila. Genome Res. 13, 2213–2219 (2003).
28. Khalturin, K., Hemmrich, G., Fraune, S., Augustin, R. & Bosch, T. C. More than just
orphans:aretaxonomically-restrictedgenesimportantinevolution?TrendsGenet.
25, 404–413 (2009).
29. Nolte,A.W.,Renaut,S.&Bernatchez,L.Divergenceingeneregulationatyounglife
history stages of whitefish (Coregonus sp.) and the emergence of genomic
isolation. BMC Evol. Biol. 9, 59 (2009).
30. Levine, M. T., Jones, C. D., Kern, A. D., Lindfors, H. A. & Begun, D. J. Novel genes
derived from noncoding DNA in Drosophila melanogaster are frequently X-linked
and show testis-biased expression. Proc. Natl Acad. Sci. USA 103, 9935–9939
(2006).
9.
Supplementary Information is linked to the online version of the paper at
www.nature.com/nature.
Acknowledgements We thank B. Walderich for providing zebrafish, A. Nolte,
E. Blohm-Sievers, B. Kleinhenz, L. Turner and J. Bryk for laboratorysupport, R. Bakaric ´
has provided the phylostratigraphic map of C. elegans, and M. Domazet-Los ˇo and
V. Dunjko have helped with statistics. L. Boell, F. Chang and A. Pozhitkov have made
suggestionsonthemanuscript.ThisworkwassupportedbyUnityThroughKnowledge
Fund (grant No. 49), Adris Foundation and funds of the Max-Planck Society.
Computational resources were provided by CSTMB and RBI (Phylostrat Cluster).
AuthorContributionsT.D.-L.conceivedthebasicideaandconductedtheexperiments;
D.T. contributed to the evaluation and interpretation of the results. Both authors wrote
the manuscript.
AuthorInformationThemicroarraydataforzebrafishweredepositedattheNCBIGene
Expression Omnibus (GEO) repository under the accession number GSE24616.
Reprints and permissions information is available at www.nature.com/reprints. The
authorsdeclarenocompetingfinancialinterests.Readersarewelcometocommenton
the online version of this article at www.nature.com/nature. Correspondence and
requests for materials should be addressed to T.D.-L. (tdomazet@irb.hr).
–0.25
–0.2
–0.15
–0.1
–0.05
0
0.05
0.1
0.15
0.2
0.25
1234567891011121314
Phylostratum
Female – male TAI difference
(–0.68)
Danio
Drosophila
Figure 4 | Comparison of differences in TAI between females and males.
Comparison across phylostrata in zebrafish (Danio) and Drosophila (see
Supplementary Fig. 4 for a plot that includes the differences between stages).
The grey shaded area designates the shared part of the phylogeny between the
two species (origin of the first cell until the last common ancestor of Bilateria,
ps1–ps7). Note that the ps14 value for Drosophila is off the scale (difference is
given in parenthesis).
RESEARCH LETTER
8 1 8 | N A T U R E | V O L 4 6 8 | 9 D E C E M B E R 2 0 1 0
Macmillan Publishers Limited. All rights reserved
©2010

Page 5

METHODS
Fish keeping and sampling conditions. Zebrafish(Daniorerio) were kept in 12l
flow-throughtanks at 26.5uC (around 60 animals per tank). For accurate staging,
fertilizedeggswerecollectedwithin15-minintervalsandincubatedinPetridishes
at 28.5uC with water changes every 2–6 h. After hatching, larvae were transferred
to1-ltanksandkeptat28.5uC.Wetook,intotal,72samplesintworeplicatesthat
correspondto60stagesacrosszebrafishontogeny(50samplesbeforethesexcould
beclearlyrecognizedplus11samplesofmalesandfemaleseach).Stagingwasdone
according to post-fertilization time. Embryos were additionally staged under the
dissecting microscope according to ref. 31 to check for the consistency of post-
fertilization timing and morphological development at standard temperature
(28.5uC). Only healthy animals that showed the expected morphological features
foragivenpost-fertilizationtimeweresampled.Eachsamplecontainedaround50
individuals until the 1day and 3 h embryo stage, 15 individuals until the 10day
larval stage, 10 individuals until the 18day larval stage, 5individuals until the
45day juvenile stage, whereas in later juvenile and adult stages we sampled males
andfemalesseparatelyandeachsamplecontained2individuals.Allsampleswere
snap frozen in liquid nitrogen and stored at 280uC until RNA extraction. To
avoidseverebiasesowingtotheexcessofunfertilizedeggs,wesqueezedeggsfrom
adult females before freezing them in liquid nitrogen.
Phylostratigraphy. A full account of phylostratigraphic analysis and theoretical
underpinnings has been presented previously16–18. The zebrafish genes of the
present study (28,546, ENSEMBL release 56) were mapped on the currently best
supported phylogeny using BLAST searches against the cleaned up and addition-
ally enriched NCBI NR database, which represents the most exhaustive set of
known proteins across all organisms. Our choice of internodes (phylostrata) in
the consensus phylogeny depended on the availability of complete annotated
genomes,reliabilityofphylogeneticrelationshipsandontheimportanceofevolu-
tionary transitions. Similarly, the data of Drosophila (13,389 genes), Anopheles
(12,457genes)andCaenorhabditiselegans(19,077genes)weremappedtothebest
supported phylogenies that represent their evolutionary lineages.
RNA isolation and microarray gene expression experiments. Total RNA was
isolated using the TRIZOL plus protocol (Invitrogen). Four-hundred nanograms
of total RNA per sample were Cy3 labelled according to the one-colour Quick
AmpLabellingKitprotocol(Agilent).LabelledcRNAswerehybridizedtoAgilent
Zebrafish (V2) Gene Expression Microarray slides (4344k) for 17h at 65uC and
washed according to the Agilent protocol. Hybridized microarray slides were
scanned using an Agilent High-Resolution Microarray Scanner.
Microarray data extraction, filtering and analysis. Raw microarray image files
were processed and quality checked by Agilent’s Feature Extraction 10.7 Image
Analysis Software.Background subtracted signal intensity values that contain cor-
rection for multiplicative surface trends (gProcessedSignal) generated by Feature
Extraction Software were used for further data analysis. Using GeneSpring micro-
arraydataanalysissoftwarewefilteredprobesthatwereflaggedasnon-uniformor
as population outlier. For every of the 72 samples we calculated average signal
intensity values over the two biological replicates. Probes (60bp) were mapped
on the Danio rerio transcripts (ENSEMBL version 54) that passed the phylostrati-
graphic analysis (see below) using CD-hit software. This procedure yielded 16,188
unique probes that collapsed to 12,892 ENSEMBL predicted genes.
Phylostratigraphically mapped genes of Drosophila were linked to available
microarray data23. This procedure yielded a data set of 3,550 genes. In a similar
fashion, phylogenetically ranked microarray data sets were obtained for C. ele-
gans25(16,832 genes) and Anopheles26(3,135 genes).
Transcriptomeageindexandstatisticalanalysis.TheTAIistheweightedmean
of phylogenetic ranks (phylostrata) and is calculated for every ontogenetic stage s
as follows:
TAIS~
P
i~1ei
n
i~1psiei
P
n
wherepsiisanintegerthatrepresentsthephylostratumofthegenei(forexample,1,
theoldest;14,theyoungest),eiisthemicroarraysignalintensityvalue(obtainedfrom
AgilentZebrafish(V2)GeneExpressionMicroarrays)ofthegeneithatactsasweight
factor and n is the total number ofgenes analysed. This way ofcalculatingthe index
givesanincreasinglystrongerweighttoyoungerphylostrata,thus compensatingfor
the fact that the older phylostrata usually harbour the larger number of genes16–18.
We chose to calculate the TAI index based on the amount of expression per
gene,ratherthanbysimplyaddingupwhetherageneisexpressedornot.Although
this latter approach would also seem feasible, it runs into a technical problem. To
saythatagivengeneisexpressedornot,onewouldhavetoimposeacutoffonthe
signalsfromthemicroarrays,whichismoreorlessarbitrary,asaweaksignalona
microarray couldbe derived froma gene with very low expression level,orfrom a
highly expressed gene that is present in a few cells only. Also, absolute quantities
are difficult to compare across microarrays and a single cutoff value would not be
appropriate (see below). In balance, we have therefore opted for the expression
level as a numerator, also because one could argue that genes that are broadly
expressed at high levels should be more relevant than specialized genes.
The TAI formula can alternatively be written as:
e1
e1ze2z:::zenzps2
Theexpressione11e21…1enrepresentsthetotalsignaloftheanalysedprobeson
the microarray, whereas the ratio ei/(e11e21…1en), which can be denoted as fi,
represents the partial concentration (frequency) of probe i in the total microarray
signal at a given stage; it is within a range between zero and one. It is important to
note that the calculation of the partial concentration (fi) inherently makes a global
intensity normalization over the microarrayexperiment ata given stage and that at
every stage the sum of partial concentrations will equal one. In many microarray
studies it is common to assess the direction of expression change (over- or under-
expression). This type of analysis requires that after the normalization procedure,
which aims to remove noise from the experiment, expression signals that are mea-
sured across experimentsstillreflect absolute number of mRNAmolecules per unit
ofbiologicalmaterial.Insuchsituations,ifglobalintensitynormalizationisapplied,
itmustbeassumedthatthetotalnumberofmRNAcopiesforallgenesonthearray
doesnotsignificantlydifferbetweenexperiments.Contrarytothiscommonapplica-
tionofmicroarrays,inourstudywearenotinterestedinthedirectionofexpression
change of particular genes. Instead, we are looking at how partial concentrations of
RNAs contribute to the overall transcriptome across stages. For this purpose it is
irrelevant which part of the transcriptome is responsible for change of the partial
concentration. Therefore in our data treatment it is not necessary to assume that
cumulative signals do not differ between experiments. This shift in perspective
greatlysimplifiestheanalysisonthescaleofthecompleteontogenybecauseabund-
ance and distribution of transcripts is commonly very different between stages.
Thus,theTAIcanbewrittenasasumofproductsbetweenpartialconcentration
and corresponding phylostratum:
TAIS~ps1
e2
e1ze2z:::zenz:::zpsn
en
e1ze2z:::zen
TAIS~
X
n
i~1
psifi~ps1f1zps2f2z:::zpsnfn
To asses the contributions of a specific phylostratum to the overall TAI (Figs 1b
and3b)wesplittheabovetotalsumofpsifiproductstosubsetsofpsifisumswhere
the value of psi(phylostratigraphic rank) was used as a grouping factor.
By applying repeated measures ANOVA on these psifiproducts we tested the
significanceofdifferenceinTAI betweenstages.RepeatedmeasuresANOVAwas
used because the same set of probes are measured at every stage, that is, there is
dependence between the stages compared. Before means of these products across
stagesarecomparedbyANOVAwemultipliedeverypsifiproductwithconstantn
(total number of analysed probes). This transformation does not influence the
ANOVA analysis andits sole purpose is that means of psifiproductscompared in
ANOVA are equal to the corresponding TAI values. Because the assumption of
sphericitywasviolatedinthedatasetsanalysedbyrepeatedmeasuresANOVA,we
appliedtheGreenhouse–Geissercorrection.Multivariateteststatistics,analterna-
tiveapproachthatisnotdependentontheassumptionofsphericity,corroborated
our statistical results of repeated measures ANOVA. We used the bootstrap
approach (1,000 replicates) to asses the standard error of weighted mean (TAI)32.
Relative expression of the genes for a given phylostratum (ps) and devel-
opmental stage (s) (Fig. 2) was calculated according to the equation:
RE(ps)s~
?f{?fmin
?fmax{?fmin
where?f is the average partial concentration of RNAs from phylostratum ps for a
given stage and?fmax,?fminarethemaximal and minimal average partialconcentra-
tion from phylostratum ps across all considered stages, respectively.
31. Kimmel, C. B.,Ballard,W. W., Kimmel, S. R., Ullmann, B. & Schilling, T. F. Stages of
embryonic development of the zebrafish. Dev. Dyn. 203, 253–310 (1995).
32. Efron, B. & Tibshirani, R. Bootstrap methods for standard errors, confidence
intervals, and other measures of statistical accuracy. Stat. Sci. 1, 54–75 (1986).
LETTER RESEARCH
Macmillan Publishers Limited. All rights reserved
©2010