|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: AAA. Performed the experiments: DAF. Analyzed the data: DAF AAA. Contributed reagents/materials/analysis tools: AAA. Wrote the paper: AAA.
Planaria continue to blossom as a model system for understanding all aspects of regeneration. They provide an opportunity to understand how the replacement of missing tissues from preexisting adult tissue is orchestrated at the molecular level. When amputated along any plane, planaria are capable of regenerating all missing tissue and rescaling all structures to the new size of the animal. Recently, rapid progress has been made in understanding the developmental pathways that control planarian regeneration. In particular Wnt/beta-catenin signaling is central in promoting posterior fates and inhibiting anterior identity. Currently the mechanisms that actively promote anterior identity remain unknown. Here, Smed-prep, encoding a TALE class homeodomain, is described as the first gene necessary for correct anterior fate and patterning during planarian regeneration. Smed-prep is expressed at high levels in the anterior portion of whole animals, and Smed-prep(RNAi) leads to loss of the whole brain during anterior regeneration, but not during lateral regeneration or homeostasis in intact worms. Expression of markers of different anterior fated cells are greatly reduced or lost in Smed-prep(RNAi) animals. We find that the ectopic anterior structures induced by abrogation of Wnt signaling also require Smed-prep to form. We use double knockdown experiments with the S. mediterranea ortholog of nou-darake (that when knocked down induces ectopic brain formation) to show that Smed-prep defines an anterior fated compartment within which stem cells are permitted to assume brain fate, but is not required directly for this differentiation process. Smed-prep is the first gene clearly implicated as being necessary for promoting anterior fate and the first homeobox gene implicated in establishing positional identity during regeneration. Together our results suggest that Smed-prep is required in stem cell progeny as they form the anterior regenerative blastema and is required for specifying anterior cell fates and correct patterning.
Understanding the genetic basis of tissue regeneration (remaking) from adult structures is an important long-term goal for biomedical science. The widespread nature of regenerative phenomena in different animals allows us to study evolution's answers to coordinating this process. We use the relatively simple and experimentally tractable planarian model to study this process. After almost any amputation these animals unerringly replace all missing tissues. This ability has two key components. Firstly, planaria have a population of stem cells capable of rapidly dividing and becoming all the cell types that are missing, such as muscle, gut, and brain cells, after amputation. Secondly, the genetic information in these stem cells and the remaining tissue is able to coordinate the regeneration process so that new structures are the correct size and in the correct place. This allows the production of a fully functional individual at the end of the regeneration process. We are specifically interested in how structures end up in the correct place in new tissue they form. Here we discover and describe the role of a gene, called Smed-prep, particularly central to this process. Smed-prep is required to coordinate the regeneration of the planarian brain, arguably the most exciting part of planarian regeneration.
Planaria continue to blossom as a model system for understanding all aspects of regeneration –. A sustained and passionate effort by a number of scientists is pushing planaria to the forefront of the regeneration field, both technically ,  and theoretically , and they are finally starting to be directly informative of phenomena in other systems . They provide an opportunity to understand how the replacement of missing tissues from preexisting adult tissue is orchestrated at the molecular level. When amputated along any plane planaria are capable of regenerating all missing tissue and rescaling all structures to the new size of the animal .
Recent work has shown that conserved signaling pathways play a role in axial patterning during both regeneration and homeostatic tissue turnover –. In particular Wnt/beta-catenin signaling is necessary for posterior fate during regeneration, with loss of beta-catenin or Wnt signaling leading to all amputations regenerating anterior structures and a gradual loss of posterior identity during homeostasis , , . Conversely, over activity of Wnt signaling induced by abrogating the expression of negative regulators of the pathways leads to ectopic posterior fate . Further studies have begun to describe the temporal nature of this posterior specification circuit, as well the conserved nature of upstream regulation , .
Previously elegantly executed manipulative work has uncovered phenomena that suggest that anterior fated tissue can inhibit the regeneration of anterior fate elsewhere . In addition some headway has been made in understanding the potential signaling systems responsible for this , . In particular the planarian nou-darake (ndk) gene, an FGF-like receptor, has been shown to be necessary to restrict the formation of anterior-dorsal brain ganglia/cephalic ganglia (CG) to anterior regions . Currently though nothing is known about the instructive signals required to promote anterior fate. We wished to uncover these signals that together must promote anterior fate and correctly pattern the brain as it reforms from stem cell progeny at anterior blastemas.
Given the involvement of conserved pathways already uncovered we hypothesized that other genetic circuits employed to specify positional domains in other animals would be responsible for this process during planarian regeneration. One obvious group of genes for this process would be planarian orthologs of the Hox genes and Hox gene co-factors, These are required for anterior-posterior axis specification in the metazoa , . Planarian Hox orthologs have been previously studied, and in some cases are expressed in distinct spatial domains, but have as yet no functions are assigned to them in planaria.
This has led us to consider TALE class homeodomain containing genes, a subset of which act as Hox gene cofactors . Collectively, they are known to modulate the activity of Hox proteins by regulating their localization within the cell and by increasing their binding site specificity, but also have many hox independent roles in development –.
Here, Smed-prep, encoding a TALE class homeodomain, is described as the first gene that is necessary to instruct anterior fate and patterning during planarian regeneration.
The Smed-prep transcript was identified in an informatics screen for homeodomain proteins in the Schmidtea mediterranea genome. Searching the S. mediterranea genome identifies other TALE class homeodomain proteins , but Smed-prep encodes the only PREP ortholog (Figure 1A). The protein encoded by Smed-prep has high homology to other PREP proteins and contains the conserved features expected of this protein family (Figure S1). In vertebrates, PREP proteins have been implicated in a number of key developmental processes , including the correct patterning of anterior structures . The function of Hox and Hox co-factors in planaria remains enigmatic. The fact that these two groups of homeodomains act together to pattern tissues in other systems makes them strong candidates for a role in providing positional information in planarians. For this reason we performed a detailed study of Smed-prep.
We performed in situ hybridization on whole and regenerating asexual planaria , . We find that Smed-prep is expressed at ubiquitously low levels throughout the parenchyma and at higher levels in the head region. The posterior margin of anterior expression coincides with the most posterior position of cephalic ganglia (CG) (Figure 1C and 1D). We also detect low levels of Smed-prep expression in the posterior midline, at higher levels than the broad parenchymal expression, in approximately 50% (39/72) of animals (Figure 1B). Smed-prep expression is not sensitive to irradiation, indicating that Smed-prep is not expressed in, or dependent on, the ‘neoblast’ stem cells (data not shown). During regeneration induced by pre- and post-pharyngeal amputation (Figure 1E) Smed-prep expression is first detected at 24 h and is present in both anterior and posterior blastemas (Figure 1F). New Smed-prep expression is not detected at 6, 12 or 18 hours of regeneration. Expression in the anterior is bilateral up to 3 days but has expanded across the whole blastema at 5 days (Figure 1G and 1H). At 5 days Smed-prep is expressed throughout the anterior compartment with the notable exception of the eye field. We also detect feint expression in the posterior midline of approximately 50% of trunk fragments at 3 (18/41 fragments) and 5 days (23/40 fragments) of regeneration. We observe this in trunk fragments only (Figure 1G and 1H). This expression is absent later and presumably reappears after regeneration is complete and animals reach a homeostatic state (see above). At 8 days of regeneration, posterior blastema expression is reduced while expression in the anterior continues to be high (Figure 1I). This expression pattern led us to hypothesize a role for Smed-prep in patterning regenerating tissue after amputation. In particular expression in whole worms suggested that Smed-prep might have a role in pattering and/or maintaining anterior structures.
We performed RNAi ,  of Smed-prep to investigate its function during regeneration (see Figure S2 for summary of injection protocols). Smed-prep dsRNA injection before inducing regeneration by amputation (Figure 1E) resulted in all worms having either a cyclops phenotype (Figure 2A) or no eyes at all (Figure 2B, Table 1). All animals had correct early blastema formation, normal levels of neoblast proliferation (data not shown) and no defects in posterior blastema formation (Figure 2A, 2B, 2D, and 2E). A similar cyclops phenotype has been described for a S. mediterranea slit ortholog . Staining with an anti-arrestin VC-1 antibody specific for planarian photoreceptors and associated neurons  we observed that the single eye phenotype appeared to represent a fusion of two eyes (Figure S2D, S2E). We detected no other midline defects in regenerating animals that were described for Smed-slit, and Smed-slit expression itself was normal (Figure S2F and S2G). This suggests, in agreement with the Smed-prep expression pattern, that the cyclops phenotype is due to a defect in anterior patterning and fate rather than any midline defects. Control gfp(RNAi) animals had normal eye structure (Figure S2E).
We investigated the structure of the planarian ventral nerve cords (VNCs) and CG using the anti-SYNORF1 (3C11) cross-reactive monoclonal antibody . We found that in all Smed-prep(RNAi) animals the CG were greatly reduced, with almost no brain at all discernible in the most severely affected RNAi worms (Figure 2D and 2E). In these animals anti-SYNORF1 positive cells do form from differentiating neoblast progeny in the anterior as part of the VNCs. Significantly, anti-SYNORF1 positive cells are present along the whole anteroposterior axis. This suggests, along with correct pharynx and posterior regeneration that Smed-prep(RNAi) does not affect the general ability of stem cells to differentiate. All control gfp(RNAi) animals were normal (Figure 2C and 2F). We confirmed the loss of CG by looking at the expression of Smed-GluR (specific for CG (Figure 2I and 2M). This loss of anterior structures suggests a role for Smed-prep in patterning anterior structures and/or a requirement for Smed-prep in allowing neoblasts to differentiate into CG cells. This phenotype is different from that previously described for the S. mediterranea ortholog of adenomatous polypolis coli (APC), a negative regulator of Wnt signaling. Smed-APC-1(RNAi) results in ectopic posterior fate at anterior blastemas .
To build a more exact picture of the requirements for Smed-prep we also investigated its role during regeneration more directly. We injected regenerating animals after amputation and then re-amputated (Figure S2). This approach has previously been used as a proxy to separate regeneration specific effects from homeostatic effects . Control gfp(RNAi) worms regenerated normally but Smed-prep(RNAi) worms failed to make eyes and CG almost entirely (Figure 2G and 2H, Table 1). All animals did regenerate normal VNCs within regenerated anterior tissue. This confirms that new Smed-prep expression during regeneration is required to properly replace anterior structures.
To investigate whether Smed-prep was required specifically for stem cell progeny to differentiate to CG or instead primarily for global anterior fates we investigated the expression of cintillo  and Smed-sFRP-1 , . These genes represent two different anterior markers that are not expressed in CG cells. We find that both cintillo and Smed-sFRP-1 expression are greatly reduced or absent in Smed-prep(RNAi) animals at 12 days of regeneration (Figure 2J and 2K). In the case of Smed-sFRP-1 expression we observed a correlation between the strength of the Smed-prep(RNAi) phenotype and whether any Smed-sFRP-1 expression was detectable. Those animals that maintained a single eye (and therefore some CG) also had some remaining Smed-sFRP-1 expression. Animals with stronger phenotypes (no eyes) had no detectable anterior Smed-sFRP-1 expression. All gfp(RNAi) animals had normal expression for both these markers (Figure 2N and 2O). Together these data suggest that Smed-prep is required for correct anterior blastema fate patterning during regeneration, rather than solely for CG formation by differentiating neoblasts.
This loss of anterior markers led us to consider whether Smed-prep(RNAi) leads to a homeotic like posteriorisation of the planarian body plan. We found no evidence for this by looking at the relative position of the regenerating or fully formed pharynx, the expression of a medial marker Smed-Tcen49 , or by looking at the expression of posterior markers such as Smed-HoxD . Thus we infer that Smed-prep(RNAi) leads to a reduction in the formation of anterior structures, but neither a change to posterior fate at anterior blastemas nor an expansion in posterior or medial fates in existing tissues (Figure S2J, S2K, S2L, and S2M). We also found that early Smed-sFRP-1 expression at anterior blastemas at 24 hours of regeneration is absent in Smed-prep(RNAi) animals. This suggests Smed-prep acts to provide anterior fate and pattern the anterior blastema, after polarity is set (Figure S2H and S2I).
The planarian brain and the planarian head have distinct A/P polarity, as is the case in other animals . Smed-prep expression is higher in the anterior and lateral margins of the planarian head (Figure 1B). We wished to know whether this was a reflection of Smed-prep having a role in defining different A/P fates within the anterior blastema itself. In this case any remaining brain fated tissues observed in Smed-prep(RNAi) animals (Table 1) would be expected to have posterior brain fate. By investigating the expression of Smed-WntA, a marker of the posterior brain  we found that Smed-prep(RNAi) animals that regenerated one eye and some CG also maintained antero-posterior identity within their much reduced anterior structures (Figure 2L and 2P). In these animals Smed-WntA still labels a posterior domain of the remaining CG. This suggests that Smed-prep is required to specify an anterior field of cells in which further A/P patterning occurs.
We performed long term Smed-prep(RNAi) in whole worms, to assess its role during normal homeostasis and tissue turnover. Long-term knockdown did not result in loss or proportional reduction of anterior structures or CG/Brain (Figure 2R, Table 1). However, Smed-prep(RNAi) worms developed a new pair of photoreceptors anterior to the original pair (Figure 2Q). This result suggests that Smed-prep expression in the anterior of whole worms is required for correct positioning of the photoreceptors during homeostasis but not for CG maintenance. Smed-sFRP-1 expression was also affected in these animals, with loss of anterior margin and lateral expression, but maintenance of weaker ventral antero-medial expression (Figure 2S). This provides more evidence to suggest that Smed-sFRP-1 expression is dependent on Smed-prep expression. These data show that Smed-prep has different roles in establishing anterior structures and their subsequent maintenance.
The finding that the CG were not reduced in homeostasis led us to consider whether Smed-prep(RNAi) would affect the lateral regeneration of anterior structures. We reasoned that if Smed-prep was not required for CG maintenance during homeostasis, then alternative anterior maintenance mechanisms must be active during homeostasis. These alternate mechanisms could also be sufficient to orchestrate lateral regeneration, a scenario where existing anterior structures are left partially intact. We cut Smed-prep(RNAi) worms longitudinally (Figure S2A) and observed regeneration. We found that Smed-prep(RNAi) worms were able to laterally regenerate all structures, with correct scaling, and subsequent normal behavior. While some worms did not regenerate a second eye correctly, all animals regenerated lateral CG. However, on looking at the pattern of the CG structure in more detail we noticed that the bilateral CG fused at the anterior tip (Figure 2T and 2U). In this regenerative scenario Smed-prep(RNAi) animals can regenerate antero-laterally but CG structures are not patterned correctly. This indicates that while Smed-prep is specifically required for the replacement of missing anterior structures when they are absent, it is not required to generate missing anterior fated structures during antero-lateral regeneration, i.e. when one side of the brain is still present. Instead, it is only required for the formation of correct pattern during this regenerative scenario. It seems likely that the remaining anterior tissue contains cues, generated downstream of Smed-prep during normal anterior regeneration, that are sufficient to direct neoblast progeny to CG fate.
Our experiments thus far suggest that Smed-prep is required for anterior patterning and fate. To formally rule out the possibility that Smed-prep is also directly required during anterior regeneration for stem cell differentiation into CG we utilized the previously described nou-darake (ndk) RNAi phenotype . RNAi of this FGF-like receptor gene leads to ectopic posterior expansion of CG during homeostasis and regeneration. We predicted that if Smed-prep was required for anterior patterning but not for neoblast differentiation then double Smed-prep/ndk(RNAi) worms would display expanded CG differentiation, but with aberrant anterior patterning and loss of anterior marker expression. Smed-prep/gfp(RNAi) and Smed-ndk/gfp(RNAi) animals regenerated with reduced and expanded CG respectively compared to gfp(RNAi) worms (Figure 3B and 3C). Smed-prep/ndk(RNAi) animals had expanded CG but this expansion was patterned incorrectly (Figure 3D). The CG of Smed-prep/Smed-ndk(RNAi) animals are fused at the anterior tip, similar to Smed-prep(RNAi) laterally regenerated animals (Figure 3D). Both gfp(RNAi) and smed-ndk/gfp(RNAi) animals have normally patterned bilateral CG (Figure 3A and 3C). To test if this mispatterning was concomitant with the loss of anterior fate we also looked at Smed-sFRP-1 expression. Whereas Smed-sFRP-1 expression was normal in Smed-ndk(RNAi) animals after regeneration it was absent or greatly reduced in Smed-prep/ndk(RNAi) animals (Figure 3E–3G). This suggests that Smed-prep specifies an anterior domain during regeneration and that stem cell progeny normally differentiate to form CG only within this domain. This restriction requires activity of Smed-ndk, which is also expressed in an anterior domain. In double Smed-prep/ndk(RNAi) animals the loss of Smed-ndk removes this restriction on neoblast progeny, allowing them to adopt CG fate without the presence of Smed-prep expression, but does not rescue the defects in anterior patterning.
Wnt signaling is central in patterning the antero-posterior axis of planarians by promoting posterior fate , , , . Given the finding that Smed-prep is not required for CG maintenance or formation during homeostasis and lateral regeneration respectively, it remained unclear whether Smed-prep would be required for the ectopic anterior structures observed when Wnt signaling is attenuated. We found that when Smed-beta-catenin-1(RNAi) results in head regeneration at both anterior and posterior blastemas –, ectopic and prolonged expression of Smed-prep in these new heads is observed (Figure 3H). In addition Smed-prep/beta-catenin-1(RNAi) reduced anterior structures at both ends (Figure 3K). As Smed-prep expression is initially present at both posterior and anterior blastemas our data suggest that active Wnt signaling in the posterior blastema suppresses Smed-prep action at posterior blastemas post-transcriptionally.
Smed-prep is the first gene clearly implicated as being necessary for promoting anterior fate during regeneration in S. mediterranea. We propose that after initial polarity determination, involving Wnt signals and other as yet unknown mechanisms, Smed-prep expression in neoblast progeny determines an anterior field of cells in which anterior structures differentiate and are patterned. At posterior blastemas Smed-prep activity is inhibited post-transcriptionally by Wnt activity. This now provides the opportunity to discover downstream genes that are required for further fine patterning during anterior regeneration, as some of these are likely transcriptional targets of Smed-prep activity.
In other animals the function of PREP TALE class homeodomains remains rather poorly defined compared to those of other TALE class family genes. In the both major invertebrate genetics models, C. elegans and D. melanogaster, a direct ortholog of PREP TALE class homeodomains is absent . Interestingly both worms and flies contain MEIS orthologs (unc-62 and homothorax respectively) that have broad roles in specifying fate during development ,  and other members of the nematode and arthropod phyla do have PREP orthologs . The finding that PREP is involved in zebra fish brain development may suggest that PREP has an evolutionary conserved role in anterior fates. Broader phylogenetic study of its function is required to test this . Here, we show that Smed-prep expression and function delineates the whole anterior domain, including all regions of the brain. Previous studies of Hox and Hox co-factor function have not implicated these two groups of genes in defining the most anterior structures of other vertebrates  or arthropods .
Significantly, the requirement for Smed-prep is observably different during homeostasis and different regenerative scenarios. This illustrates that the genetic networks available to solve different regenerative scenarios may be diverse and are likely to depend on the informational/signaling capacity of the differentiated portion of starting tissue. In addition it is the first time that homeobox transcription factors have been directly implicated in A/P patterning in planaria. We suspect that other conserved homeodomain proteins will also play core roles in specifying positional information during regeneration.
All experiments were performed with a clonal line originally generated from a single animal of the asexual strain of the planarian S. mediterranea collected in Montjuïc (provided by Professor Emili Saló i Boix) maintained at 20°C in tap water treated with activated charcoal and buffered with 0.5 ml/L 1 M NaHCO3. Planarians were fed veal liver and starved for at least one week prior to experiments.
To identify planarian homologues of TALE transcription factors we searched a local database of Version 3.1 of the S. mediterranea Genome Project for orthologs of mammalian TALE genes (http://genome.wustl.edu/genomes). The contigs 018898 and 020093 containing Smed-prep were analyzed using Vector NTI (Invitrogen) and sequence data supplemented by using RACE (Ambion RLM Race Kit). The primers Sm-Prep-Forward with sequence ATTGCTACTAGAGCAATGTGAACAAGC and Sm-Prep-Reverse with sequence ATTCTGCGTCGGGCATTGAT amplify a 810 bp fragment which was used for whole mount ISH hybridization and RNAi knockdown. PREP and TALE proteins sequences were taken from Mukherjee at al  and alignments checked with the CLUSTAL . Phylogenetic reconstruction was conducted using MEGA version 4 using the bootstrapped neighbour-joining method . The Smed-prep sequence has been submitted to GenBank with accession number GU290186.
DsRNAs were synthesized as described previously . Control animals were injected with dsRNA of GFP that has no homology in the planarian genome. DsRNA microinjection was performed as described elsewhere . For injection schedules please refer to Figure S2. For double RNAi experiments concentrations for each gene were maintained at 1 µg/µl after mixing and for GFP controls 2 µg/µl was injected.
Whole mount ISH hybridization was carried out as described previously  with modifications described in  and . The paraformaldehyde solution for the fixation step was prepared fresh and adjusted to pH 9.5 using NaOH.
For immuno-staining animals were killed in 2% HCl for 5 min on ice and then fixed in Carnoy's solution for 2 h at 4°C. After fixation, samples were processed as described elsewhere , . The following primary antibodies were used: anti-SYNORF1, a mouse monoclonal antibody specific for synapsin (Developmental Studies HybridomaBank, dilution of 125) and anti-arrestin VC-1, a mouse monoclonal antibody specific for planarian photosensitive cells (kindly provided by Hidefumi Orii, used at a dilution of 115,000). Goat anti-mouse secondary antibody conjugated to Alexa 488 or Alexa 546 (Molecular Probes) was used at a 1400 dilution.
Brightfield pictures were taken on a Zeiss Discovery V8 from CarlZeiss using an AxioCam MRC from CarlZeiss. Fluorescent pictures were taken on a Leica MZ16F fluorescence stereomicroscope using a Leica DFC 300Fx camera (Leica Lasertechnik, Heidelberg). Confocal laser scanning microscopy was performed with a LeicaSP2 confocal laser scanning microscope (CLSM) (Leica Lasertechnik, Heidelberg).
Alignment of Smed-prep translation to other animal PREP proteins. Alignment of Smed-Prep across the conserved MEIS and Homeodomain regions of this TALE class protein with other animals. The Smed-Prep translation is underlined in red.
(0.03 MB PDF)
RNAi protocols and characterization of Smed-prep function. Explanation of RNAi injection schemes and further analysis of Smed-prep function. Figurative explanation of RNAi injection and amputation protocols used for assaying Smed-prep function. In the standard protocol animals receive 3×32 nl injections of dsRNA at 1 μg/μl for three consecutive days before pre- and post- pharyngeal or longitudinal amputations are performed (A). To assay the effect of Smed-prep(RNAi) specifically during regeneration animals tails are amputated and injected 3 times with 3×32 nl injections of dsRNA at 1 μg/μl as depicted. The animals are then re-amputated (B). Homeostasis experiments were conducted for 28 days or longer. Initially animals were injected as in (A) but instead of being amputated they were left intact, fed and injected with a single set of 3×32 nl injections of dsRNA at 1 μg/μl for the subsequent weeks. Staining with the anti-arrestin VC-1 monoclonal antibody against the photoreceptor neurons shows that Smed-prep(RNAi) animals have only one photoreceptor, which appears to be a fusion of two normal eyes (D). gfp(RNAi) animals always regenerate a normal visual system (E). The midline of Smed-prep(RNAi) animals (G) seems normal and Smed-slit expression that labels cells in the midline of gfp(RNAi) animals (F) is unaffected. The expression of Smed-sFRP-1 appears early during anterior regeneration. At 24 hours of regeneration it can already be seen in the blastema in gfp(RNAi) animals (H). In Smed-prep(RNAi) animals expression is not detected in tail pieces even when the sample is left to develop until background is very high (I). The expression of HoxD is detected in the tail parenchyma up to the mouth of the pharynx, in the mouth itself and in a few scattered cells just anterior to the pharynx in gfp(RNAi) animals (J). There is no ectopic expression detected in the head of Smed-prep(RNAi) animals (K). The normal expression domain of Smed-Tcen49 in scattered cell clusters in the trunk region of the planaria (L) is not expanded anteriorly in Smed-prep(RNAi) animals (M).
(0.52 MB PDF)
We thank laboratory members for comments on the manuscript and encouragement. We are grateful to C. González-Estévez, E. Saló, F. Cebrià, P. Newmark, T. Adell, D. Molina, D. Evans, and M. Iglesias for their collaboration and sharing reagents.
The authors have declared that no competing interests exist.
Work in AAA's laboratory is supported by project grants from the BBRSC (www.bbsrc.ac.uk) and MRC (www.mrc.ac.uk). AAA received personal support from a Wellcome Trust International Research Fellowship and now a Research Councils of the United Kingdom Fellowship during the course of this work. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.