One of the main questions in developmental biology is whether embryos develop in a mosaic (Roux,1888) or a regulative manner (Driesch,1892). In mosaic patterning, a manipulated embryo region will give rise only to those regions it would normally have produced in the undisturbed situation. In contrast, in regulative patterning, an abnormal complement of cells in an embryo (or embryo part) has the potential to generate a normal embryo (or embryo part; after Driesch,1892). Although developmental processes most often lie somewhere between perfect mosaicism and perfect regulation (for a historic summary, see Lawrence and Levine,2006), it is still difficult to ascertain the regulative capacities of a particular system, and some intense debates remain unsolved (see below).
The developing limb possesses several advantages that make it a particularly useful model for addressing these questions, especially in relation to tissue manipulation. First, the limb is dispensable for survival, which permits every kind of so-called cut and paste experiments in the chick embryo. Second, the outcome of every manipulation done in one limb can be easily compared with the normal outcome observed in the contralateral limb. Finally, fate maps are available for all developmental stages, and the size and shape of the different elements at every stage is very well known. Moreover, the molecular signals and genetic networks that operate within the limb are also active in other territories of the embryo, allowing the extrapolation of conclusions.
The vertebrate limb bud or primordium arises from the lateral plate as a bulge of mesenchymal cells encased within an ectodermal hull. Two main signaling centers are known to govern growth and pattern along its main axes. The apical ectodermal ridge (AER), a thickened epithelium that rims the distal end of the bud, is necessary for mesenchymal cell survival and proliferation, maintenance of an undifferentiated state, and proximodistal (P-D) elongation (Saunders,1948; Rowe and Fallon,1981; Rowe et al.,1982; Dudley et al.,2002). The zone of polarizing activity (ZPA), located at the posterior region of the limb mesenchyme, is essential for sustained proliferation and anterior–posterior (A-P) patterning (Saunders and Gasseling,1968; Tickle et al.,1975; Towers et al.,2008). Several members of the fibroblast growth factor (FGF) family are produced by and can substitute for the AER (Niswander et al.,1993; Fallon et al.,1994). The ZPA produces the diffusible molecule sonic hedgehog (SHH), which mediates its mitogenic and patterning activities (Riddle et al.,1993; Chiang et al.,1996; Towers et al.,2008; Zhu et al.,2008). Other diffusible molecules are also known to affect the limb, such as retinoic acid (RA) produced by the territory proximal to the limb bud (Swindell et al.,1999). These and other signals might be involved in the limb positional information system, and are thus the usual suspects analyzed in studies of regulative processes in the limb. RA and FGFs play antagonistic roles during P-D patterning, the former as an inducer of proximal character (Mercader et al.,2000; Yashiro et al.,2004), and the latter as a distalizing signal (Mercader et al.,2000; Mariani et al.,2008). However, the debate about the regulative abilities of the limb bud (especially along the P-D axis) started long before these molecules were discovered.
The first reports on regulative development in the limb bud came from the “French School.” Two types of assays were performed: regulation of tandem duplications (Kieny,1964b; Amprino and Camosso,1965) and regulation of interstitial deletions along the P-D axis. For the purposes of the present study, we shall focus on the latter. In a series of experiments aimed at demonstrating that P-D inductive power resided in the mesenchyme, several researchers showed that after elimination of an intermediate slice from the early limb bud, a complete or almost complete P-D axis was produced quite frequently (Wolff and Hampe,1954; Hampe,1959; Kieny,1964a,c; Kieny and Pautou,1977). In similar and independent experiments, Amprino and Camosso also observed some regulation, although it was described as “imperfect” (Amprino and Camosso,1958,1959a,b).
The other side of the coin is represented by the “English School.” Summerbell, Lewis and Wolpert performed heterochronic transplantation of distal limb bud tips, leading to both supernumerary and deficient primordia (based on the available fate maps). They reported a complete lack of regulation (Summerbell et al.,1973; Summerbell and Lewis,1975); i.e., the composite limbs developed according to the fate of the transplanted parts, in agreement with their cell-autonomous theory for P-D patterning (the so-called “progress zone” model; Summerbell et al.,1973). However, when the analysis was extended to cover a wider range of conditions, a considerable degree of qualitative and quantitative regulation was reported for the early stages of limb development, before any differentiation had occurred (Summerbell,1977,1981).
The regulative capacities of the developing limb were also tested in a somewhat different kind of experiment, in which the distal tip of the limb primordium was grafted to different locations, such as the somites, the embryo flank or the proximal limb bud, and the outcome scored (Zwilling,1956; Amprino and Camosso,1959a,b). The results, however, seemed to vary significantly depending on the size and the exact location of the graft site. When very thin “apical slivers” were grafted to the flank, it was reported that the whole P-D axis was produced, instead of just part of the autopod (Zwilling,1956). However, with thicker grafts no such P-D regulation was observed, regardless of the grafting site (somites, flank or proximal wing bud; Amprino and Camosso,1959a,b). Therefore, the effect of graft size seems an interesting issue to revisit.
Regarding the mechanism that could lead to regulation in cases where it was observed, two obvious possibilities come to mind: some of the distal cells are “reprogrammed” by the proximal mesenchyme toward more proximal fates, and/or some proximal cells acquire more distal fates due to the influence of the distal graft. The first attempts to distinguish between these possibilities were made by placing carbon particles at the host/graft interface, and scoring their positions at the end of the experiment. Based on this approach, Kieny and co-workers claimed that most of the newly generated segment originated from the distal graft, suggesting proximalization of the distal limb bud (Kieny,1964a,c). However, this system has been argued to be unreliable (Lewis,1975), casting doubt on the validity of the results. Another approach was to make leg-to-wing or wing-to-leg grafts, and rely on the morphology of the regulated element to infer its origin (Kieny,1964c). This method is also unreliable, however, because on some occasions wing tissue can be incorporated into leg-shaped elements (Kieny,1964c; Krabbenhoft and Fallon,1989) and vice versa (Saunders et al.,1955; Krabbenhoft and Fallon,1989). To try and clarify this issue, Kieny and co-workers made quail/chick combinations, which permitted histological determination of tissue origin (Kieny and Pautou,1977). In these experiments, the intermediate element was shown to contain contributions both from the proximal tissue (mainly at the preaxial region) and from the distal tissue (mainly at the postaxial region). Strikingly, regulation was only observed in the chick-to-quail experiments and not in the reverse experiments, raising the possibility that differences in growth rate or other parameters between the two species could affect the results and thus limit their significance.
Regarding the mechanism of this putative proximalization of fate, ectopic application of RA has been shown to produce a proximalizing effect on the expression of P-D molecular markers (Mercader et al.,2000) and cell adhesion properties (Tamura et al.,1997; Mercader et al.,2000), making RA one of the main putative regulatory signals. However, there are also studies in the literature showing distalizing effects. For example, limb cells are recruited toward more distal fates under AER influence (Saunders et al.,1955); a new limb bud is generated when FGFs are applied to the early interlimb flank (Cohn et al.,1995); and indeed the whole P-D axis is regenerated when, after distal limb bud amputation, the proximal stump is covered with an AER (Zwilling,1956; Hampe,1959) or treated with FGF-soaked beads (Taylor et al.,1994; Kostakopoulou et al.,1996,1997). This recruitment of cells under AER influence does not seem to affect the somites, however (Amprino and Camosso,1958,1959a,b). Finally, classic and recent studies have shown that small groups of cells from very early limb buds can change their P-D anatomical contribution and molecular identity when heterotopically grafted from proximal to distal regions or vice versa (Krabbenhoft and Fallon,1989; Wyngaarden and Hopyan,2008). Proximal tissue looses this plasticity earlier than distal tissue (Wyngaarden and Hopyan,2008), suggesting that only undifferentiated tissue has this reprogramming capacity.
Despite the significant advances made over the decades, researchers have encountered two persistent problems when studying the regulative abilities of the limb bud. First, interpretation of results as evidence for regulation or mosaicism has depended on the accuracy of the available fate maps, because it is from these that researchers deduced which prospective regions were being removed or grafted. Any inaccuracy or misinterpretation of the fate maps could lead to an erroneous interpretation of the results. To circumvent this problem, Summerbell used a strictly quantitative approach, calculating a mathematical function that correlated the P-D extension of the operated primordium with the P-D length of the resulting wing (Summerbell,1977). Any deviation from the line expected for perfect mosaic development was interpreted as evidence of at least partial regulation. Despite its originality, this approach had a major disadvantage: it assumed uniform expansion along the P-D axis. It is now known that this is not the case, at least for the developing mouse limb (Boehm et al.,2010), while less clear evidence exists for the chicken limb (Hornbruch and Wolpert,1970; Fernandez-Teran et al.,2006). The second problem is the limited ability to recognize which elements are formed from the graft and which from the host. As summarized above, all systematic approaches to achieving this identification have been somewhat flawed, except where wing- or leg-specific markers were used to trace tissue origin in wing/leg combinations (Dudley et al.,2002). Unfortunately, the phenomenon of regulation was not thoroughly studied in this latter case.
Here, we have addressed the question of regulative vs mosaic limb development by means of transplantation of distal limb bud tips to different regions of the embryo. The comparison of the different outcomes, combined with the molecular tracing of tissue origin, indicated that the regulation observed mainly involves recruitment and distalization of host tissue, rather than proximalization of grafted cells.
To evaluate the regulatory capacities of distal limb bud cells, we transplanted undifferentiated distal tips from chick limb buds of various stages to different host locations. In contrast to the majority of similar experiments in previous reports (Zwilling,1956; Amprino and Camosso,1959a,b; Dudley et al.,2002), we decided to tilt the cutting plane posteriorly to compensate for the posterior shift that has been described especially in the fate maps of forelimb (FL) buds but also hindlimb (HL) buds (Vargesson et al.,1997). This procedure enabled us to take cells with more homogeneous and distally restricted fates than is achieved by cutting perpendicular to the P-D axis (Fig. 1A). For the site of implantation, we chose two regions of the limb bud—the proximal region of Hamburger and Hamilton stage 20 limb buds (HH20; Hamburger and Hamilton,1992; Fig. 1B, blue arrow) and HH24 distal wing buds (prospective zeugopod; Fig. 1B, red arrow)—and two regions outside the limb bud—HH20 anterior hindbrain (Fig. 1B, red arrow, pooled with grafts to prospective zeugopod) and HH20 paraxial mesoderm (hereafter somites, Fig. 1B, green arrow). Instead of relying on the published fate maps to detect possible regulative effects, we used a comparative method. One of the distal tips obtained from the FL/HL buds of each donor embryo was grafted to a “proximal” site (somites or proximal limb; experimental graft), while the other tip was grafted to a “nonproximal” one (prospective zeugopod or anterior hindbrain; control graft). See the Experimental Procedures section for a detailed description.
Three categories of donor embryos were used: HH19–HH20−, HH21–HH22 and HH24. The effect of the grafting site was semiquantitatively scored by subdividing the P-D axis of the graft according to the presence of the following structures as the proximal-most element (see examples in Fig. 2A): the proximal epiphysis of the stylopod (S1) or the zeugopod (Z1); the diaphysis of one of these segments (S2 or Z2); the distal epiphysis (S3 or Z3); the whole tarsometatarsus (A1); only part of the tarsometatarsus (A2); or only the phalanges (A3).
We performed some grafts of distal wing bud tips to somites and zeugopod/hindbrain to validate the system (data not shown), confirming that the mean value of the segments obtained correlated well with the fate map of the transplanted regions (Saunders,1948; Stark and Searls,1973; Bowen et al.,1989; Vargesson et al.,1997; Sato et al.,2007). However, because we sometimes needed to perform heterotopic wing/leg transplantations and grafting to the HH24 wing bud is easier than to the leg bud, we soon switched to using leg bud donors only. After the complete series of experiments involving 200-micron thick grafts (n = 150), in the vast majority of the cases we found no significant differences between grafts transplanted to the different locations (data not shown).
This independence of graft identity from graft site agrees with previous reports indicating autonomy of distal limb grafts (Amprino and Camosso,1958,1959a,b; Summerbell et al.,1973; Summerbell and Lewis,1975), but disagrees with others that reported regulation depending on the stage and size of the transplant or the way the operation was done (Hampe,1959; Kieny,1964a,c; Kieny and Pautou,1977; Summerbell,1977,1981). To test the influence of graft size, we repeated the experiments using 100-micron thick distal limb bud tips (n = 85). Grafts to somites and zeugopod/hindbrain formed mainly autopod structures: A1 from HH19–HH20−, A1–A2 from HH21–HH22 and A2–A3 from HH24 donors (Fig. 2B and left panels in Fig. 3A–C). In the case of HH21–HH24 donors, these structures were similar to those formed by the 200-micron grafts to the same sites (data not shown), suggesting that the distal 200 microns of the limb bud are homogenous in P-D potential during stages 21–24. On the other hand, grafts to proximal wing buds tended to form elements of a more proximal character (mainly zeugopod, see below) than those formed by grafts to zeugopod/hindbrain or to somites (Figs. 2B, 3), suggesting nonautonomous regulation. Moreover, there was a match between the most proximal structures formed by the graft and the P-D level of the graft site. Interestingly, most of the grafts to the proximal wing bud that did not show this apparent regulation had taken very superficially, far from the endogenous skeletal elements (data not shown). Because the apparent regulation was observed only in transplants to proximal limb, these findings suggest two possible explanations: proximalization of the graft or recruitment and distalization of host limb cells.
To determine the origin of the newly generated structures that did not correspond to the original fate of either the host or the grafted tissue, we examined leg-to-wing transplants for the expression of the leg marker Tbx4 and the wing marker Tbx5 (Chapman et al.,1996; Gibson-Brown et al.,1996). The results indicate that the proximal segments acquired by regulation are derived entirely from the host in all cases analyzed (n = 19), with no intermingling of cells from different origins (Fig. 4A,B,D,D′). This was corroborated by the pattern of ectodermal derivatives covering the acquired structures, which matched that of the host limb (Fig. 4B, arrowheads).
A very important aspect of this experiment is whether the element recruited from the host incorporates cells originally fated to the stylopod (which would undergo recruitment and distalization) or only those fated to the zeugopod (which would just be recruited, not instructed). Gross examination of whole limbs or stained cartilaginous elements provided some insight into this subject. Of 27 grafts that formed zeugopodal structures, 14 (51.9%) were fused or in close proximity to the host stylopod (Fig. 3B, arrow, and Fig. 4A), 6 (22.2%) were fused or close to the ulna (Fig. 3A,C, 4B) and 7 (25.9%) were found at the elbow joint (not shown), suggesting that frequently host cells are not only recruited, but also instructed. To further confirm this, we thoroughly marked the proximal wing wound with the lipophylic dye DiI (1,1′, di-octadecyl-3,3,3′,3′,-tetramethylindo-carbocyanine perchlorate) before performing the transplantation (Fig. 4E). Two or three days after grafting, the distribution of the fluorescent signal was examined. In 93% of the specimens (n = 13/14), the dye was detected in the host stylopod but not in the zeugopod (Fig. 4F). Of those cases, nine specimens (69%) showed a strong fluorescent signal in the proximal element of the ectopic limb, confirming that quite frequently this structure was derived from the host stylopod (Fig. 4F,H,H′). In the remaining cases (n = 4/13), no strong regulation was found, as judged by very restricted dye incorporation into the ectopic limb (data not shown). Finally, expression of the zeugopod marker Hoxa11 was analyzed in six of the specimens in which stylopod-fated host cells had been recruited to the hybrid structure, revealing that the recruited element expressed this marker (Fig. 4G, n = 6/6). Taken together, the results indicate that in all cases the zeugopod of the ectopic limb arises from the host, excluding proximalization of graft tissue as the origin of this structure. In addition, in a majority of cases, the new zeugopod derives from stylopod-fated host cells, indicating both recruitment and regulation of host cells toward more distal (zeugopod) fates. In the remaining cases, the new zeugopod presumably derives only from the host prospective zeugopod, indicating only recruitment and not regulation of host cells' fate. It is noteworthy that when the thin grafts were transplanted to the distal rather than the proximal limb region, recruitment of host tissue was sometimes observed as well (Fig. 4C, n = 2/4); however, as shown above, this never gave rise to the formation of skeletal elements more proximal than those normally derived from the grafted fragment (Fig. 2B).
Here, we have tested the ability of cells of the chick posterodistal limb bud tip to regulate or reprogram their normal fate in response to different external cues provided by graft sites. For example, the somites and proximal limb bud contain high endogenous levels of the putative proximalizing signal RA (Rossant et al.,1991; Maden et al.,1998), whereas the prospective HH24 zeugopod and anterior hindbrain contain lower RA levels. Our grafting system is very similar to others used in earlier studies (Zwilling,1956; Amprino and Camosso,1958; Amprino and Camosso,1959a,b; Dudley et al.,2002), but with a subtle difference. We tilted the cutting plane posteriorly to transplant only the distal-most fates. This ensures that we always took part of the ZPA, something that was very unlikely in earlier studies, especially those involving very thin grafts. We think that this innovation might improve the growth and development of the elements produced by our grafts, rendering the results easier to interpret. An additional difference is that we used a comparative system, thus avoiding reliance on published fate maps when scoring regulation. As previously shown (Amprino and Camosso,1958; Amprino and Camosso,1959a,b; Summerbell et al.,1973; Summerbell and Lewis,1975), when thick (200-micron) distal limb tips were transplanted, no major effect was observed, and development proceeded according to the original fate of the transplanted cells (data not shown). However, a different outcome was observed with thin (100-micron) distal tips. Whereas grafts transplanted to the somites (green arrow in Fig. 1B) behaved in the same way as controls (red arrow), grafts transplanted to the proximal wing (blue arrow) formed more proximal elements than controls in 55–100% of cases, depending on the donor stage (Figs. 2B, 3). Molecular characterization and tracing of tissue origin revealed that the intermediate segment in the regulated specimens was a zeugopod composed entirely of host cells, most of which belonged, originally, to the prospective host stylopod (Fig. 4). These results, in addition to the reported recruiting and distalizing abilities of the AER and FGFs (Zwilling,1956; Hampe,1959; Taylor et al.,1994; Kostakopoulou et al.,1996,1997; Mercader et al.,2000; Mariani et al.,2008), suggest that when the graft is sufficiently thin the signals produced by its AER can reach, recruit, and instruct host cells to generate a hybrid host/graft limb. The fact that grafts that took very superficially did not form any extra element suggests that pre-chondrogenic mesenchyme needs to be specifically recruited to the formation of the new elements. These results suggest that the maximum distance over which AER signals can diffuse to reach host tissue lies somewhere between 100 and 200 microns. Further experiments with grafts of different thickness showed that the maximum distance that supports reproducible regulation is roughly 170 microns (data not shown). This distance is very similar to the 162 microns that has been reported to be the limit of AER influence on small fragments of leg mesoderm grafted to subapical regions of the wing bud (Krabbenhoft and Fallon,1989).
Competence to respond to distalizing signals seems to be restricted to limb cells (this study and Amprino and Camosso,1959b; Taylor et al.,1994; Kostakopoulou et al.,1996,1997) or limb-competent flank regions (Zwilling,1956; Cohn et al.,1995), and is not found in somites (this study and Amprino and Camosso,1958; Amprino and Camosso,1959a,b). In addition, the new element recruited by the graft's AER always matches the missing fates between the graft site and the graft itself (i.e., a zeugopod is formed when an autopod-fated graft is attached to the host elbow). This is quite reminiscent of the mechanism that operates during vertebrate limb regeneration (Maden,1980).
The recruitment and distalization mechanism we have described could explain evidence for regulative development obtained in classical experiments in which the origin of the newly formed region was not specifically assessed (Zwilling,1956; Summerbell,1977; Summerbell,1981). However, the absence of graft proximalization conflicts with the previous observations of the contribution of distal tissue to the intercalated structures developed after chick-to-quail transplants (Kieny and Pautou,1977). One possible explanation is that the fate maps used to interpret those earlier results were not sufficiently accurate (see Summerbell and Lewis,1975), suggesting that the presumptive autopod graft included the postaxial part of the prospective chick zeugopod, while the preaxial part of the resulting chimeric zeugopod originated from the proximal quail tissue. The question of why the reverse combination does not lead to regulation remains, however, unsolved. Nevertheless, we speculate that the quail AER might be unable to generate FGF signals above a threshold concentration needed to recruit chick cells.
Finally, our results could also be applied to recent models on P-D patterning. It has recently been proposed that P-D specification depends on an intercalative mechanism (Mariani et al.,2008). According to this model, the proximal and distal elements are specified first by means of the opposing actions of proximal and distal signals, and subsequently, the intermediate element arises as a consequence of the interaction between the preexisting segments (Mariani et al.,2008). Our results, based on the experimental apposition of proximal and distal fates, seem to suggest that an intercalative mechanism is possible, however the following arguments suggest the contrary. Within the distal 200 microns, the limb bud cells of several stages are homogeneous in P-D fate potential after transplantation. Therefore, if intercalation of proximal and distal fates was the cause of the generation of the intermediate fates, no differences should have been found between the 100- and the 200-micron transplants to the proximal wing. On the contrary, the distance to the AER would be an essential parameter in the case that the signals emanating from it were responsible for the regulation observed. Therefore, it seems more likely that such regulation is based on signals rather than on cell-to-cell interactions and thus it would not be intercalation in the classic sense (French et al.,1976). In addition, the intercalation model suggests that during limb development AER signals only induce the distal-most fates, with intermediate ones being generated by apposition of proximal and distal fates. However, our experiments strongly suggest that AER signals induce zeugopodal fates from the prospective stylopod. This scenario favors a progressive mechanism, in which the zeugopod precedes the autopod in the sequence of specification.
In conclusion, our results demonstrate that regulative limb development can occur through the recruitment of proximally fated cells likely influenced by AER distalizing signals.
Fertilized eggs were obtained from Gibert farm (Tarragona, Spain) and stored at 16°C until use. To continue development, eggs were incubated at 38°C in a nonrotating incubator. The day before manipulation, eggs were windowed as described (Ros et al.,2000) and embryos were staged according to Hamburger and Hamilton (Hamburger and Hamilton,1992). Incubation time was then fine-tuned to obtain the desired stages on the day of the experiment. Only donors of the correct stage were used for experiments.
On the day of operation, 100-micron- or 200-micron-thick slivers were cut (straight line) from stage 19–20−, 21–22, or 24 limb buds in an uncoated Petri dish containing sterile phosphate-buffered saline (PBS). A calibrated reticule was used to measure the distance from the AER. At the magnification used, each subdivision equaled 17 microns. The thickness was accurate probably within the 1-subdivision range and surely within the 2-subdivision range (66–134 microns in the case of 100-microns grafts). The cutting plane was tilted slightly posteriorly with respect to the plane orthogonal to the proximodistal axis (see main text). Both contralateral limb buds were cut, and the slivers compared in the dish and, if necessary, one was trimmed to match the size of the other. Every effort was made to ensure that both distal tips from the same donor were as comparable as possible, especially regarding thickness. In the cases where this was not achieved, the thinner graft was grafted to the “proximal” region (somites or proximal wing bud), to exclude the possibility that production of extra elements in the experimental graft was due the presence of excess tissue from the outset.
Graft sites were prepared by making small rectangular wounds (or a single straight line in case of the hindbrain) in the host tissue with a tungsten needle, and the grafts were pipetted onto the wound and pushed in with a pair of blunt forceps. No staples or pins were used to keep the grafts in place. After transplantation, embryos were kept at room temperature for 1 hr before being returned to the incubator. The PBS used to keep the embryos moist contained penicillin and streptomycin (Gibco) at 10 times the normal working concentration for cell culture; three–four drops were used per host. Embryos were collected 2 to 7 days after transplantation and processed for in situ hybridization or cartilage staining.
DiI Labeling Experiments
Stock and working solutions of DiI were prepared as described (Sato et al.,2007). A mouth pipette and a pulled capillary tube were used to apply the dye to the embryo.
Standard procedures were used for staining with Victoria Blue or Alcian Green (Bryant and Iten,1974; Morgan et al.,1992).
In Situ Hybridization
Embryos were fixed overnight at 4°C in 4% paraformaldehyde (PFA) in PBS. Whole-mount in situ hybridization was performed as described (Wilkinson and Nieto,1993), except for an added bleaching step with 6% hydrogen peroxide, and the use of BM purple (Roche) instead of nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate (NBT/BCIP) to develop the stain. Proteinase K (PK) was diluted in PBS containing 0.1% Tween 20 (PBST); HH26 or HH35 limbs were treated with 25 or 60 μg/ml PK, respectively, during 25 min at room temperature. Whole-mount stained specimens were sometimes embedded in paraffin and microtome-sectioned (8–12 microns thick).
We thank all the members of the laboratory for insightful discussions, and Simon Bartlett for editing. We thank anonymous reviewers for suggestions that helped improving the study. A.R.-D. is the recipient of a research contract from Madrid's regional government (CAM). M.T. was funded by the Spanish Ministry of Science and Innovation (MICINN). The CNIC is supported by the MICINN and the Pro-CNIC Foundation.