The periodic coloration in birds forms through a prepattern of somite origin

How birds change their stripes From stripes to spots, animals often exhibit periodic coloration. Discrete embryonic domains (prepatterns) precede the periodic feather patterns observed in birds. After documenting natural variation in the striped pattern of galliform birds, Haupaix et al. performed long-term skin grafts to transfer the pattern of one species to another (see the Perspective by Prud'homme and Gompel). This approach revealed that periodic stripe formation obeys developmental landmarks upstream of local refining mechanisms. The somitic mesoderm first instructs stripe position through the early expression of the pigmentation gene agouti, which then controls stripe width by modulating pigment production in a dose-dependent manner. Thus, during feather patterning, a two-step process is at play. Science, this issue p. eaar4777; see also p. 1202 The somite instructs a prepattern of colored stripes in galliform birds. INTRODUCTION In animals, coat color is often arranged in periodic motifs that vary widely, from striped to spotted patterns. These intricate designs have long fascinated developmental biologists and mathematicians alike. What are the mechanisms underlying the formation of periodic patterns and shaping their diversity? Spatial organization in the developing skin involves prepatterns that precede the color pattern. Self-organizing events have long been thought to act upstream of prepatterns (e.g., through molecular diffusion or pigment cell interaction). Changes in both of these molecular and cellular events may contribute to periodic pattern variation. However, periodic patterns are highly reproducible within species and display specific orientation and periodicity, which suggests that they also rely on preexisting spatial reference. RATIONALE Documenting phenotype diversity constitutes a promising framework for the prediction of such spatial landmarks, comparable to mathematical modeling strategies. We surveyed variation in the transient periodic pattern visible in juvenile birds of the galliform group, in which longitudinal stripes are organized in a black-yellow-black sequence in the dorsal region. RESULTS By comparing the striped pattern for 10 galliform bird species, we showed that the width of each stripe varied and that their number increased with dorsum size. In contrast, their absolute positions were comparable. We analyzed pigment appearance in the embryonic skin of five representative species and showed that the periodic striped pattern results from the timely production of yellow coloration at specific locations. This yellow-production pulse was not triggered by a certain stage of feather growth or by dynamics of feather follicle production across the dorsum. However, it was linked to the early expression of agouti. This well-known pigmentation gene displayed a composite expression pattern in longitudinal bands whose width and position correlated with that of yellow stripes in each species. To test agouti’s role, we used a functional approach by exploiting mutant strains of quails: The increase (in the Fawn strain) or decrease (in the recessive black strain) of agouti expression levels respectively led to wider or narrower yellow stripes. Comparing pigment distribution across feathers between these gain- or loss-of-function mutants and wild-type quails showed that agouti controls stripe width by adjusting the duration of the yellow-production pulse in a dose-dependent manner. Both the position of agouti-expressing bands and that of yellow stripes did not change in mutant quails. To identify the origin of signals controlling localized agouti expression and setting the position of yellow stripes, we performed heterospecific grafting experiments: Embryonic tissues from donor quails were transplanted into pheasant hosts. We found that after transplanting somites (from which dermal cells originate), chimeras locally displayed quail-like expression of agouti in the developing skin. Long-term experiments showed that hosts displayed a striped color pattern typical of the donor at the level of the graft. Such changes were not observed when the neural tube (from which pigment-producing cells originate) was grafted. These results showed that the somitic mesoderm autonomously instructs agouti expression and consequently the position of yellow stripes. CONCLUSION We conclude from this work that the galliform striped pattern is achieved in a two-step mechanism. The somite provides positional information to the developing dermis; this controls the position of agouti expression in a prepattern that foreshadows yellow stripes. Their width is then refined by agouti, which locally controls yellow production in a dose-dependent manner. This sequential organization of space, combining early landmarks and local mechanisms, may govern the formation (and thus constrain the evolution) of many periodic patterns. The striped pattern of a Japanese quail embryo. Galliform birds display a longitudinal pattern of colored stripes already visible a few days before hatching (here in a Japanese quail, Coturnix japonica). Stripes form through differential deposition of black and yellow pigments along growing feathers in the dorsum. Our work shows that this pattern is controlled by a prepattern instructed by the somitic mesoderm. The periodic stripes and spots that often adorn animals’ coats have been largely viewed as self-organizing patterns, forming through dynamics such as Turing’s reaction-diffusion within the developing skin. Whether preexisting positional information also contributes to the periodicity and orientation of these patterns has, however, remained unclear. We used natural variation in colored stripes of juvenile galliform birds to show that stripes form in a two-step process. Autonomous signaling from the somite sets stripe position by forming a composite prepattern marked by the expression profile of agouti. Subsequently, agouti regulates stripe width through dose-dependent control of local pigment production. These results reveal that early developmental landmarks can shape periodic patterns upstream of late local dynamics, and thus constrain their evolution.

INTRODUCTION: In animals, coat color is often arranged in periodic motifs that vary widely, from striped to spotted patterns. These intricate designs have long fascinated developmental biologists and mathematicians alike. What are the mechanisms underlying the formation of periodic patterns and shaping their diversity? Spatial organization in the developing skin involves prepatterns that precede the color pattern. Selforganizing events have long been thought to act upstream of prepatterns (e.g., through molecular diffusion or pigment cell interaction). Changes in both of these molecular and cellular events may contribute to periodic pattern variation. However, periodic patterns are highly reproducible within species and display specific orientation and periodicity, which suggests that they also rely on preexisting spatial reference. RATIONALE: Documenting phenotype diversity constitutes a promising framework for the prediction of such spatial landmarks, comparable to mathematical modeling strategies. We surveyed variation in the transient periodic pattern visible in juvenile birds of the galliform group, in which longitudinal stripes are organized in a black-yellow-black sequence in the dorsal region.
RESULTS: By comparing the striped pattern for 10 galliform bird species, we showed that the width of each stripe varied and that their number increased with dorsum size. In contrast, their absolute positions were comparable. We analyzed pigment appearance in the embryonic skin of five representative species and showed that the periodic striped pattern results from the timely production of yellow coloration at specific locations. This yellow-production pulse was not triggered by a certain stage of feather growth or by dynamics of feather follicle production across the dorsum. However, it was linked to the early expression of agouti. This well-known pigmentation gene displayed a composite expression pattern in longitudinal bands whose width and position correlated with that of yellow stripes in each species. To test agouti's role, we used a functional approach by exploiting mutant strains of quails: The increase (in the Fawn strain) or decrease (in the recessive black strain) of agouti expression levels respectively led to wider or narrower yellow stripes. Comparing pigment distribution across feathers between these gain-or loss-of-function mutants and wild-type quails showed that agouti controls stripe width by adjusting the duration of the yellowproduction pulse in a dosedependent manner. Both the position of agoutiexpressing bands and that of yellow stripes did not change in mutant quails.
To identify the origin of signals controlling localized agouti expression and setting the position of yellow stripes, we performed heterospecific grafting experiments: Embryonic tissues from donor quails were transplanted into pheasant hosts. We found that after transplanting somites (from which dermal cells originate), chimeras locally displayed quail-like expression of agouti in the developing skin. Long-term experiments showed that hosts displayed a striped color pattern typical of the donor at the level of the graft. Such changes were not observed when the neural tube (from which pigmentproducing cells originate) was grafted. These results showed that the somitic mesoderm autonomously instructs agouti expression and consequently the position of yellow stripes.
CONCLUSION: We conclude from this work that the galliform striped pattern is achieved in a two-step mechanism. The somite provides positional information to the developing dermis; this controls the position of agouti expression in a prepattern that foreshadows yellow stripes. Their width is then refined by agouti, which locally controls yellow production in a dose-dependent manner. This sequential organization of space, combining early landmarks and local mechanisms, may govern the formation (and thus constrain the evolu- The striped pattern of a Japanese quail embryo. Galliform birds display a longitudinal pattern of colored stripes already visible a few days before hatching (here in a Japanese quail, Coturnix japonica). Stripes form through differential deposition of black and yellow pigments along growing feathers in the dorsum. Our work shows that this pattern is controlled by a prepattern instructed by the somitic mesoderm. The periodic stripes and spots that often adorn animals' coats have been largely viewed as self-organizing patterns, forming through dynamics such as Turing's reaction-diffusion within the developing skin. Whether preexisting positional information also contributes to the periodicity and orientation of these patterns has, however, remained unclear. We used natural variation in colored stripes of juvenile galliform birds to show that stripes form in a two-step process. Autonomous signaling from the somite sets stripe position by forming a composite prepattern marked by the expression profile of agouti. Subsequently, agouti regulates stripe width through dose-dependent control of local pigment production. These results reveal that early developmental landmarks can shape periodic patterns upstream of late local dynamics, and thus constrain their evolution.
M any vertebrates display intricate color patterns characterized by a periodic arrangement of pigments in stripes or spots (1). This spatial organization was recently shown to involve genes whose embryonic expression profile forms a prepattern preceding the adult pattern [e.g., the developmental expression of the pigmentation genes edn3b and alx3 foreshadows striped patterns in the fur of cats and African striped mice, respectively (2, 3)]. A longstanding challenge has been to uncover the em-bryonic pattern-forming events acting upstream of prepatterns to create discrete compartments in the developing skin. Computer simulations of stochastic dynamics such as Turing's reactiondiffusion, which involves the interaction of at least one self-activating factor and its inhibitor diffusing at long range, reproduce periodic motifs that resemble those observed in the wild; this finding suggests that color self-organizes in the skin tissue [see (4) for a review]. This hypothesis is supported by a handful of empirical studies. In zebrafish, longitudinal stripes form through interaction [depolarization-repulsion (5)] of pigment cells sequentially aggregating locally, forming interstripes, then expanding and compacting within stripe regions (6,7). In striped cats, the aminopeptidase taqpep creates periodicity by establishing an edn3b-expressing prepattern (2). Changes in cell behaviors or in the biological parameters of molecular players (e.g., clearance rate, which reflects the rate of molecule elimination; diffusivity) may contribute to natural variation in periodic patterns, consistent with theoretical predictions: Modifying the corresponding parameters of simulations gives rise to a vast array of patterns (4). Most periodic color patterns, however, display specific orientation or periodicity and are highly reproducible within species, which suggests that their formation does not entirely rely on stochastic events. Here, we investigated whether early developmental landmarks provide preexisting spatial reference to periodic patterns.

Stripes vary in number and width but have comparable positions relative to the dorsal midline
To predict potential spatial landmarks, we surveyed phenotypic variation in the transient striped pattern visible along the dorsum of juvenile birds of the galliform group (8,9). In flat-skin specimens for 10 species chosen for their representative varying patterns (table S1), we compared feather types according to pigmentation along barbs (i.e., primary branches). We distinguished two main types: (A) Flat-skin preparations of dorsal skin regions (white rectangle) in juvenile individuals of 10 galliform bird species (n = 2 to 5 per species, see table S1) show their alternating yellow and black stripes, labeled as cy (central yellow), cb (central black), ly (lateral yellow), lb (lateral black), and ly2 (additional lateral yellow) stripes. This pattern is symmetrical over the dorsal midline (dm, in red) and was thus used as a landmark reference. Scale bars, 1 cm. (B) Measures of absolute distances from the dorsal midline to the boundary between each stripe at wing and leg levels [indicated by Wi and Le in (A) and (C)] demonstrate interspecies variation in stripe width. We also observed changes in the number of lateral stripes despite differences in dorsum size (species are shown in size order, with smallest at top). Conversely, cb and ly stripes have comparable positions. Green lines represent cb distance from the dorsal midline (2 to 4 mm at wing level, 2 to 3 mm at leg level); blue lines represent ly distance from the dorsal midline (3 to 9 mm at wing level, 3 to 6 mm at leg level). C.j, Coturnix japonica; O.p, Oreortyx picta; P.p, Perdix perdix; A.r, Alectoris rufa; N.m, Numida meleagris; C.am, Chrysolophus amherstiae; S.r, Syrmaticus reevesii; P.c, Phasianus colchicus; C.au, Chrosoptylon auritum; M.g, Meleagris gallopavo. Error bars indicate SD. (C) Feathers of the dorsal tract organize in feather rows (fr; shown in orange in C. japonica). Their spatial arrangement varies between species. Plotting feather types at each position of the tract produces precise spatial representations of the color pattern. black (b; entirely eumelanic) and yellow (y; with a pheomelanin band and black base and tip) ( fig. S1). This allowed us to identify a common stripe sequence, symmetrical over the dorsal midline and extending from the wing to the tail: The most central stripe is black (cb)-sometimes containing a few central yellow feathers (cy)-and is flanked by two lateral yellow stripes (ly). The latter can be ventrally bordered by lateral black stripes (lb) and additional lateral yellow stripes (ly2) (Fig. 1A). We observed that the number of stripes often increased with dorsum size, hence variation in color pattern does not result solely from scaling. The position of each stripe relative to dorsum size varied between species ( fig. S2). In comparison, absolute distances between the dorsal midline and the center of the stripes closest to the median (i.e., cb and ly) were comparable; this suggests that these stripes are positioned early, prior to dorsal skin expansion (Fig.  1B). The width of each stripe also varied between species. To detail this variation, we compared stripe boundaries relative to feather tracts, which are feather-covered skin areas separated by nearglabrous regions. Within tracts, feathers form longitudinal rows whose number, spacing, and length along the anteroposterior axis is typical to each species (10, 11) ( fig. S3). We plotted feather types at each position of the dorsal tract and found that color boundaries are highly reproducible within species' tracts (e.g., in Coturnix japonica, the cb-ly boundary is characterized by the consistent production of b/y split feather types). However, these color boundaries vary between species relative to feather row number (i.e., stripe width) and along the anteroposterior axis (i.e., stripe shape; Fig. 1C). This suggests that variation in stripe width or shape results from differences in local mechanisms occurring during feather tract formation.
Interspecies differences in agouti expression correlate with stripe pattern variation We took advantage of both the absence of variation in stripe position and the presence of differences in stripe width or shape to link the juvenile color pattern to (i) pigment production and (ii) tract formation during embryogenesis (in contrast to phenotypic surveys classically limited to observations of adult patterns). We used five species representative of variation in the width or shape of stripes, the presence or absence of cy/lb stripes, and the organization of the dorsal tract (namely C. japonica, Alectoris rufa, Perdix perdix, Phasianus colchicus, and Syrmaticus reevesii). We first compared the appearance of pigments in these species. We found that pigment production starts a few days before hatching (when all feather follicles of the dorsal tract are visible; fig. S4) and occurs in a medial-tolateral wave: Eumelanin pigments are first produced in the most central feathers, forming one (P. colchicus, S. reevesii) or two (C. japonica, A. rufa, P. perdix) longitudinal black bands, and then progressively appearing laterally. This forms the cb stripe. Eumelanin production is transiently 2 of 6 replaced in lateral feather rows by pheomelanin, forming the banded pattern in feathers of ly stripes. The most lateral feathers produce only eumelanin (forming lb stripes), and the most central feathers transiently produce pheomelanin (forming the cy stripe; Fig. 2A). Thus, the striped pattern results from localized "pulses" of yellow pigment production (cy and ly stripes) in an otherwise black-producing tissue.
To test whether yellow pulses occur at certain stages of tract differentiation, we reconstructed the sequence resulting in the spatial arrangement of follicles across the dorsal tract in all species and compared it with the position of the yellowproducing domains. Developing follicles stained with the b-catenin marker (12) first correspond to one (P. colchicus, S. reevesii) or two (C. japonica, A. rufa, P. perdix) longitudinal bands; rows of follicles are then added in a wave traveling ventrally [consistent with previous observations in the chick (13)], and lastly in the most central area, completing the dorsal tract ( fig. S5). Tracing back cy and ly feather rows to their order of for-mation showed that this order differs between species (Fig. 2B). Thus, the position of cy and ly stripes is not set by a mechanism taking place at given stages of tract formation.
We next tested whether yellow pulses are triggered by feather growth: We compared, in the ly stripe, (i) stages at which feathers switch to pheomelanin or back to eumelanin production, and (ii) the size and proximodistal position of the pheomelanin band in fully grown y feathers (as a readout for the onset and duration of the pulse). We found no correlation between these and the timing of yellow production ( fig. S6). Thus, yellow pulses are not temporally linked to stages of feather growth. Because yellowproducing domains have shared positions and varying widths that do not correlate with dynamics of tract formation, we predicted that they rely on prepatterns established before skin differentiation. The signaling peptide Agouti controls the production of pheomelanin pigments in a timely fashion in the hair (14) and marks prepatterns of light color domains in the embryonic skin (3,15) and along individual feathers (16). We found that prior to follicle formation in the five bird species, two agouti-expressing bands (A1) form in the dermis on both sides of the neural tube. In C. japonica, A. rufa, P. perdix, and weakly in S. reevesii, a thinner central band of agouti (A2) also appears a few hours after A1, and in C. japonica only, additional thin bands (A1′) intercalate between the A1 and A2 bands, a few hours after A2 appearance (Fig. 2C and fig. S7). Thus, agouti expression is composite in space and time. In all species, the pattern of agouti-expressing bands resembled that of yellow stripes.
Agouti foreshadows yellow stripes and controls their width in a dose-dependent manner To confirm this spatial correlation, we linked the position of each band to presumptive feather rows using double in situ hybridization for agouti and b-catenin in C. japonica. We found that the A1′ agouti band is expressed in the presumptive domain of the second feather row, consisting of 3 of 6  fig. S6; upper graph) when they first switch to yellow production (yellow dot plots) or back to black production (black dot plots) show that the yellow pulse occurs earlier in RB -/than in wild-type individuals. Measures of the size of the yellow band relative to the length of fully grown feathers ( fig. S6 and see scheme in C. japonica; lower graph) show that the yellow pulse occurs for a longer duration in Fa -/than in wild-type individuals. Error bars indicate SD. the first b/y feather of the ly stripe, whereas A1 corresponds to the third through fifth rows, forming the rest of the ly stripe, and A2 covers the central rows that form the cy stripe. In this species, agouti is later restricted to feather follicles of these stripes (Fig. 3A). Thus, the composite expression of the (A1′)A1 and A2 agouti bands reveals a prepattern that foreshadows the position of ly and cy yellow stripes, respectively. To understand how agouti expression affects stripe patterning, we used a functional quantitative approach and assessed the striped phenotype in mutant C. japonica strains. In recessive-black quails (RB -/-), agouti's spatial pattern is maintained with no change in the position or width/shape of the A1, A1′, and A2 bands ( fig. S8), but a frameshift mutation in agouti's coding sequence results in a marked decrease in transcript expression levels (17). We found that these birds display overall shorter feathers organized in thinner ly stripes (i.e., only feather rows 3 and 4 produce yellow, versus rows 2 and 5 in wild-type individuals), and no visible cy stripe (Fig. 3B). These changes in ly stripe width result from a decrease in the duration of the yellow pulse: Relative to wild-type quails, feathers of RB -/quails switch back to eumelanin production earlier (consequently producing fewer y feathers containing smaller yellow bands; Fig. 3C). Conversely, quails of the Fawn mutant strain (Fa -/-) are homozygous for a duplication at agouti responsible for an increase in its expression levels (17,18). In Fa -/birds, the spatial pattern of agouti remains unchanged ( fig. S8) but the ly (rows 1 or 2 through row 6) and cy stripes are wider, with y feathers displaying longer yellow bands because of a delayed switch to eumelanin production (i.e., increase in the duration of the yellow pulse; Fig. 3, B and C). These loss-or gain-of-function experiments show a role for agouti: At given positions (i.e., expressing bands), this gene regulates stripe width through a dose-dependent control of the duration of the yellow pulse. Agouti is a peptide diffusing in a paracrine manner; its localized expression may thus create a signaling gradient to which dermal cells respond according to their position: Certain thresholds of agouti levels trigger yellow production to locally fine-tune the border of yellow stripes and modulate the length of the yellow band (consistent with the latter being increased in feathers at the center of ly stripes relative to those at the edge; fig. S1). This raises the appealing possibility that striped pattern evolution is governed by differential regulation of agouti expression levels.
The somitic mesoderm controls localized agouti expression and yellow stripe position Contrary to stripe width, the position of stripes is conserved between species; this suggests that early developmental landmarks establish agouti's composite prepattern. We studied the earliest visible expression of agouti in the embryonic skin in all five species and found it comparably located in the dermis above the neural tube (A2; fig. S9) and the dorsomedial part of differentiating somites (A1; Fig. 4A). Previous quailchick grafting experiments showed that both the neural tube and somites contribute to the skin lineage. Feather follicle cells originate from the somite dermomyotome, and their spatial distribution is mesoderm-dependent (13,19). Conversely, pigment-producing cells (melanocytes) derive from the neural crest (20,21), but despite evidence 4 of 6 show that in all species, A1 bands (purple arrowheads) locate in the embryonic dermal area above the developing somite. nt, neural tube; s, somite. (B) C. japonica's three last formed somites at HH13 (i.e., immediately posterior to the wing bud) were grafted in place of those of P. colchicus; 3 days after grafting, immunohistochemistry for QCPN (in green) revealed that C. japonica's cells invaded the host tissues (n = 8). High-magnification views with DAPI stains (in blue) show that QCPN + cells locate in the developing dermis but not in the epidermis in regions corresponding to A1/A1′ (red square 2) and A2 (red square 1). At that stage, grafted regions in chimeras strongly express agouti in position of the A1 band (purple arrowhead, n = 7), similar to C. japonica but contrary to P. colchicus embryos. Eleven days after grafting, QCPN + cells also locate in the dermis, including the papillae of feather follicles, but not in the epidermis. Relative to C. japonica and P. colchicus controls (left panels, n = 3 per species), feathers of the chimera at the same stage (right panels, n = 4) resemble those of the host except in the grafted area (in green). (C) The chimera's grafted side (third series from top) has the same stripe and feather pattern (see #fr) as C. japonica (first series and color pattern diagram); conversely, its ungrafted side (fourth series) has the same stripe and feather pattern as P. colchicus (second series and color pattern diagram). In chimeras, the first feather (#fr1) is common to the grafted and ungrafted sides. n, notochord; end, endoderm. that their differentiation responds to cues from the feather papillae (22), the means by which they govern color patterning remained unclear, as grafted strains were homogeneously colored (and quail melanocytes produced black pigments in the host). We thus performed somite and neural tube transplantations from C. japonica donors into P. colchicus hosts (because both species display a typical striped pattern; Fig. 4B). Three days after transplanting the right half of the C. japonica neural tube (prior to major migration of the neural crest) in place of that of P. colchicus at the equivalent stage, no expression of agouti (normally visible in C. japonica but not P. colchicus embryos at that stage; fig. S7) was detected in the dermis of chimeras. C. japonica cells (stained with the specific QCPN marker) were observed in the developing neural tube, dermis, and epidermis. The same was true 11 days after grafting [i.e., embryonic day 14 (E14), when coloration is visible in lateral stripes]; at that stage, some QCPN + cells coexpressed the melanocyte marker Trp1 (fig. S10). In these chimeras, the grafted and ungrafted parts of the skin displayed similar color (and feather) patterns ( fig. S11). This demonstrates that the neural tube, despite generating melanocytes producing black pigments by default (20,21), does not spatially control the expression of agouti expression or the activity of melanocytes (and thus the position of yellow-producing domains). Conversely, 3 days after replacing the undifferentiated somites of P. colchicus located below wing level by those of C. japonica, QCPN + cells invaded the whole dermal compartment and we observed strong expression of agouti in the grafted dermis in the position of the A1 band (Fig. 4B). Eleven days after transplantation, QCPN + cells did not express Trp1 (fig. S12). Growing feathers of the grafted side were longer than those of the ungrafted side, spatially organized according to C. japonica's dorsal tract [fig. S11 and consistent with previous work (13)], and displayed the color pattern seen below wing level in C. japonica (note that it differs from leg level; see Fig. 2B). Specifically, lb feathers were present in the lateral part of the tract and yellow pigments were produced in feather rows 2 to 4, thereby reducing the width of the ly stripe in the grafted side relative to the ungrafted side (Fig. 4C). Thus, the localized expression of agouti and production of yellow pigments (and therefore, the striped coloration pattern) are autonomously instructed by the somitic mesoderm.

Discussion
Results from this study show that the striped pattern is achieved in a two-step mechanism. The somite procures positional information, creating specific compartments in the dermis at precise locations (controlling the position of agoutiexpressing bands and consequently yellow stripes). This prepattern is then refined according to expression levels of agouti (temporally controlling pigment-type production and consequently the width of each stripe). These results raise the possibility that most natural patterns, including periodic designs, rely on (and are constrained by) a stepwise organization of space that combines late, local events producing periodicity and early positional sources ensuring reproducibility, which is the key to fitness and proper directionality (in this case, longitudinal stripes depend on information from axial structures). Our work thus opens new avenues following from current theoretical models: Simulations of Turing's reaction-diffusion (and other self-organizing systems) can be framed using initial conditions corresponding to developmental landmarks. Extending empirical work on other natural patterns will better define such initial conditions and thus shed light on developmental constraints to color pattern evolution.

Specimen sampling
Previously euthanized specimens of 1-to 3-dayold hatchlings (for 10 species of galliform birds; table S1) and fertilized eggs (for C. japonica, A. rufa, P. perdix, P. colchicus, and S. reevesii) were obtained from authorized local breeders (Ferme de Chanteloup, Caringa, Mr. Bouly de Lesdain, Les Bois de Vaux) and stored at -20°C prior to processing. After egg incubation in Brinsea ovaeasy 190 incubators, embryos at various stages of development were dissected in PBS and fixed in 4% formaldehyde.

Stripe pattern analyses
Flat skin specimens were prepared from frozen carcasses through an incision of the skin along the ventral midline and cautious separation of the scalp from body muscles. Dissected skins were stretched flat, left to dry for 1 week, and imaged. Measures of distances between stripe boundaries were performed from the dorsal midline to the wings and legs (Fig. 1A) using a Vernier caliper (C. japonica and A. rufa, n = 5; Oreortyx picta and Chrosoptylon auritum, n = 2; all other species, n = 3). Within the measured area, all species displayed cb, ly, lb (Chrysolophus amherstiae, and S. reevesii only at leg level), and ly2 (except S. reevesii) stripes. Only C. japonica, A. rufa, P. perdix, and Meleagris gallopavo had cy stripes. Spatial reference maps of dorsal tracts were obtained by (i) plucking out all feathers (using forceps), which revealed their respective positions along longitudinal rows (i.e., feather rows) from the neck to the tail, and (ii) recording color at each position (coded yellow when feathers displayed pheomelanin and black when they only contained eumelanin).

Pigments, tracts, and feather spacing analyses
Embryonic flat-skin specimens were obtained by performing longitudinal openings from neck to tail in embryos of C. japonica, A. rufa, P. perdix, P. colchicus, and S. reevesii and separating the developing skin tissue from the body (using microscissors and forceps). Dissected skins were fixed in 4% formaldehyde and stored in 80% glycerol. Developing feathers were plucked, mounted on glass slides in fluoromount (Southern Biotech), imaged, and measured using Fiji software (23).

Expression analyses
In situ hybridization experiments were performed as described (24) using antisense riboprobes synthesized from vectors containing a 269-bp fragment of C. japonica, A. rufa, or S. reevesii's coding sequences for agouti, or an 881-bp fragment of C. japonica's coding sequence for b-catenin. For double in situ hybridizations, riboprobes were labeled with digoxigenin or fluorescein and sequentially revealed with anti-digoxigenin-AP or anti-fluorescein-AP antibodies (both 1:2000, Roche) and NBT/BCIP (Promega) or fast-red (Abcam) substrates.

Heterospecific grafting
Quail-pheasant grafting procedures were adapted from previous quail-chick grafting experiments (25): At HH13 in C. japonica, the three last formed somites (just posterior to the wing bud), or the right half of the neural tube alongside these somites, were ablated using glass microneedles and cleaned of potential additional neighboring tissue in PBS. Transplants were immediately transferred into HH13 P. colchicus hosts in which equivalent tissues had been previously removed (see schemes in Fig. 4 and fig. S10). Grafted eggs were kept humid by adding ampicillin-containing PBS solution on the yolk surface, closed with tape, and reincubated (37°C, 50% humidity) for 3 or 11 days (i.e., a stage corresponding to E14 in P. colchicus). Chimeras were dissected (feathers were rapidly removed for analysis) and fixed in 4% formaldehyde. The extent of the graft was determined by screening chimeras for the presence of (i) pigments in the dermis and epidermis, which are normally absent in P. colchicus but visible in C. japonica individuals at that stage ( fig. S11), and (ii) QCPN + cells on transverse sections.

Imaging
Whole-mount flat skins and stained/grafted embryos were imaged using an AF-S Micro NIKKOR 60-mm f/2.8G ED macro-lens equipped with a D5300 camera (Nikon) and a MZ FLIII stereomicroscope (Leica) equipped with a DFC 450C camera (Leica). Tissue sections were imaged using a BX53 fluorescence microscope (Olympus; for in situ hybridization or fluorescent stains) or a CSU-W1 spinning-disk confocal microscope (Zeiss) equipped with a CMOS flash 4 camera (Hamamatsu; for Trp1/QCPN fluorescent stains).