Tracing the origins of centrioles, cilia, and flagella

The microtubule (MT) cytoskeleton is the main component of macromolecular machineries, such as the mitotic spindle and centrioles/basal bodies (CBBs). CBBs nucleate formation of the centrosome—the primary microtubule-organizing center (MTOC) in animal cells that is involved in cell proliferation, migration, and polarity—and of cilia/flagella, which are essential for motility and sensing/responding to environmental cues (Fig. 1; Bettencourt-Dias and Glover, 2007; Nigg and Raff, 2009). CBBs are cylinders generally constructed from nine microtubule triplets organized in a ninefold symmetrical configuration reinforced by a cartwheel-like structure (Fig. 1, A and B). The canonical animal centrosome is usually composed of a mature mother and a daughter centriole (Fig. 1, A and B). Mature centrioles may acquire additional specialized structures, such as subdistal and distal appendages, involved in cytoplasmic MT anchoring and positioning the CBB within the cell (Fig. 1 A; Bornens, 2002; Azimzadeh and Marshall, 2010). Centrioles, then called basal bodies, anchor at the cell membrane to initiate the growth of the axoneme, the MT core of cilia and flagella (Fig. 1 C). Basal body differentiation may include the addition of structures such as striated rootlets, basal feet, and transition fibers, linking the basal body to the cytoskeleton and orienting it in relation to other organelles and other cells (Fig. 1, A and C). The axoneme assembles through the extension of the two most inner MTs of the basal body and is usually encased in a specialized membrane. Cilia may also contain a pair of central MTs, inner and outer dynein arms, and radial spokes that drive and regulate ciliary motility (Fig. 1 C; Satir and Christensen, 2007; Gerdes et al., 2009).

CBBs can assemble adjacent to other centrioles or de novo in the absence of preexisting structures (Fig. 1 B). In most eukaryotic cell types, cilia biogenesis, integrity, and function rely on a bi-directional transport mechanism along axonemal MTs, called intraflagellar transport (IFT; Fig. 1 C; Kozminski et al., 1993; Marshall and Rosenbaum, 2001; Engel et al., 2009). In most organisms, anterograde (base to tip) IFT is driven by heterotrimeric kinesin-II, whereas the retrograde (tip to base) movement is mediated by dynein heavy chain (Fig. 1 C; Davis et al., 2006).

Eukaryotes evolved more than two billion years ago, but we lack a fossil record from which we can track the evolution of the structure and function of subcellular organelles (Cavalier-Smith, 2010; El Albani et al., 2010). We can only infer details of early eukaryotic history using features of extant organisms. We use the general parsimony principle that the most likely evolutionary scenario is the one that requires the least amount of events (see for example Ouzounis (2005) for a discussion of ancestral genome reconstruction). So, if we observe that a given structure or gene exists in several eukaryotic groups, we assume that it emerged before their divergence, in the last eukaryotic common ancestor (LECA), and reject a scenario of convergent evolution.

The recent sequencing and annotation of eukaryotic genomes, combined with advancements in RNAi and mass spectrometry, has enabled the characterization of the composition and function of CBBs and associated structures in several eukaryotic groups (Figs. 2 and 3; Wigge et al., 1998; Ostrowski et al., 2002; Andersen et al., 2003; Keller et al., 2005; Pazour et al., 2005; Smith et al., 2005; Broadhead et al., 2006; Cao et al., 2006; Reinders et al., 2006; Kilburn et al., 2007; Liu et al., 2007; Baker et al., 2008a,b; Arnaiz et al., 2009; Boesger et al., 2009; Mayer et al., 2009). These data have allowed the use of comparative genomics in the study of CBBs and derived structures (Avidor-Reiss et al., 2004; Li et al., 2004; Branche et al., 2006). A strong correlation between the presence of a given set of genes and a structure in a wide number of species points to an involvement of those genes in the biogenesis/function of those structures (Figs. 2 and 3; Carvalho-Santos et al., 2010). This also includes organisms for which no functional studies have been performed. Moreover, these phylogenetic profiles can suggest paths for the evolution of CBBs and cilia.

Both the CBB and axoneme are remarkably conserved structures found in all major eukaryotic groups, which suggests their presence in the LECA (Fig. 2; Table S1; Cavalier-Smith, 2002). Organisms in most taxonomic groups show a ninefold symmetric CBB with cartwheel and MT triplets and an axoneme composed of nine MT doublets bound to outer and inner dynein arms and radial spokes, surrounding a central MT pair (Fig. 2). A similar structure is thus likely to have been present in the LECA (Fig. 4).

Other structures associated with the CBB, such as appendages (transition fibers/distal appendages, subdistal or other lateral appendages, basal feet or connecting fibers, and non-MT–based rootlets), may have also been present in the LECA (Figs. 2 and 4). It has been suggested that the association of CBBs with other cytoplasmic cytoskeleton elements, mediated by those appendages, is an ancestral feature ensuring cell polarization relative to its motile/sensory apparatus and the propagation of cell geometry through cell division (Azimzadeh and Marshall, 2010). For instance, in Chlamydomonas, the basal bodies anchor a set of stable MT bundles that are involved in division plane determination and thus impact the cell geometry of daughter cells. Basal bodies from Paramecium, Trichomonas, Chitrids, and choanoflagellates also show an association with cytoplasmic microtubules either directly or through an MTOC. Unfortunately, the little information available about the morphology and composition of these associated structures does not allow us, at the time, to discriminate between scenarios of common ancestry or convergent evolution.

Finally, it is likely that early flagella performed both motile and sensory functions, as these features are observed in multiple branches of the eukaryotic tree of life (Fig. 2; Satir et al., 2008; Bloodgood, 2010).

Three different hypotheses have been proposed for the presence of the CBBs and cilia/flagella in the LECA: endosymbiotic, viral, and autogenous origin. The first proposes that the cilium originated from a permanent symbiosis between an ancient Spirochete bacterium and an Archaebacterium (Sagan, 1967). Satir et al. (2007) proposed that the CBB evolved from an infection with a virus bearing a ninefold symmetry, whose capsid gave rise to a cartwheel with the ability to elongate MTs. To date, these hypotheses have had little morphological and molecular support (Satir et al., 2008).

The third hypothesis, first described by Pickett-Heaps (1974), proposes an autogenous origin of the CBB/cilium from an MTOC that organized cytoplasmic and mitotic MTs. The ninefold symmetry of these structures would have been the result of an evolutionary optimization of cilium motility (Cavalier-Smith, 1978; Mitchell, 2004; Satir et al., 2008). Although remote homologues of tubulin exist in Prokaryotes (Löwe and Amos, 2009), recent data suggest that most CBB components originated in eukaryotes: (a) no bacterial, archaeal, or viral counterparts of CBB components were found using defined conserved regions from proteins of the CBB assembly pathway (Carvalho-Santos et al., 2010; unpublished data); and (b) many CBB/cilium components such as tubulins, microtubule motors, and the IFT machinery are eukaryotic duplicates of cytoplasmic proteins (Fig. 3; Dutcher, 2003; Azimzadeh and Bornens, 2004; Jékely and Arendt, 2006; Wickstead and Gull, 2007). These results support the scenario of a eukaryotic/autogenous origin of the CBB, which we discuss in the next section.

Recent studies on the molecular mechanisms involved in the assembly and function of cilia and flagella reveal that the ninefold symmetrical CBB and axoneme is universally dependent on molecules such as SAS6, SAS4/CPAP, and BLD10/CEP135 (Fig. 1 B; Carvalho-Santos et al., 2010; Hodges et al., 2010). On the other hand, although it is not clear whether there is an ancestral mechanism for association of CBBs with the cytoplasmic cytoskeleton as discussed above, at least one component required for CBB migration and attachment to the actin cytoskeleton, MKS1, is ancestral (Fig. 3; Keller et al., 2005; Dawe et al., 2007, 2009).

The assembly of the axoneme and localization of receptors to the ciliary membrane was shown to be dependent on polarized vesicular trafficking and on the IFT machinery (Fig. 1 C and Fig. 4, B, B′, C, C′, and E; Nachury et al., 2010). Cargo destined for the cilium is assembled and sorted at the Golgi and delivered at the flagellar base. This process involves essential regulators of polarized transport, such as Rab8 and the exocyst and sorting complexes such as the BBSome (Figs. 3 and 4; Peränen et al., 1996; Moritz et al., 2001; Koumandou et al., 2007; Nachury et al., 2007; Yoshimura et al., 2007; Pereira-Leal, 2008; Jin et al., 2010). Cilia compartmentalization may be dependent on different mechanisms, as large proteins do not freely enter the ciliary lumen (Nachury et al., 2010). Recently, septins, GTP-binding proteins involved in compartmentalization, were implicated in the definition of a ciliary-specialized membrane compartment at the base of the cilium (Fig. 3; Fig. 4, B and E; Hu et al., 2010; Kim et al., 2010). Also, a Ran-GTP gradient may selectively control which molecules enter the cilium, similar to the mechanism that regulates transport through the nuclear pore (Dishinger et al., 2010). Although still poorly understood, this system may be crucial for transport into the cilium and for shuttling components between the cilium and the nucleus, hence regulating transcription in cilia-mediated signaling events.

Finally, the motility of the axoneme is dependent on a variety of protein complexes, such as dynein arms (Fig. 1 C and Fig. 3; Satir and Christensen, 2007).

A tubulin-like cytoskeleton, membrane compartmentalization with coat-like molecules, and an internal genome-containing compartment are all found in bacteria, and in particular in a recently characterized superphylum of bacteria suggested to be ancestors of eukaryotes, which includes the Planctomycetes, Verrucomicrobia, and Chlamydiae (Devos and Reynaud, 2010). Ran GTPase, which controls nucleo-cytoplasmic shuttling, appears to have emerged before the radiation of the Rab family (Colicelli, 2004). Thus, many of the essential components required to assemble a cilium, either sensory or motile, can be traced back to the proto-eukaryote.

Several unicellular organisms use their flagellum to attach to surfaces and pull the cell body in a process called gliding, which allows cells to move in an almost amoeboid fashion. A simple form of motility such as this could have been the first proto-cilium–based motility (Satir et al., 2008). This is supported by the fact that this type of movement is found in at least three different eukaryotic groups (Plants, Discicristates, and Cercozoa; Bloodgood, 1981; Kozminski et al., 1993; Saito et al., 2003; Chantangsi et al., 2008; Satir et al., 2008; Cavalier-Smith et al., 2009).

Several proteins had to emerge in early eukaryotes to allow the evolution of a motile structure with ninefold symmetry, formed by triplets and doublets of closed and opened MTs. These probably include ε- and δ-tubulins, allowing for the construction of the unique opened MTs of the CBB and cilium, and other proteins such as SAS6, SAS4/CPAP, BLD10/CEP135, POC1, and CEP164 involved in cartwheel, centriolar MTs, and appendage assembly (Figs. 1–4; Carvalho-Santos et al., 2010; Hodges et al., 2010; Azimzadeh and Marshall, 2010). In addition, multiple rounds of dynein duplication with subfunctionalization originated the outer and inner dynein arms, which produce the flagellar beating force (Wickstead and Gull, 2007). The emergence of a variety of molecules needed for biogenesis and orientation of the central MT pair would have allowed the evolution of more complex beat waveforms (Fig. 4, D and E). The ninefold symmetry of the axoneme appears to be the most efficient conformation to enclose the central pair (Mitchell, 2004, 2007; Satir et al., 2008). Whether the CBB appeared first, at the same time, or after the axoneme is not clear, but morphological distinctions between the two structures (doublet vs. triplet MTs, central MT pair, etc.) were likely present in the LECA (Figs. 2 and 4).

The evolution of a sensory cilium required the targeted trafficking of receptors and signaling molecules to a specialized membrane domain. Jékely and Arendt (2006)recently proposed an evolutionary scenario in which the initial structure to appear during CBB/cilia evolution was an individualized sensory patch resulting from polarized vesicle transport from the Golgi to the plasma membrane using specific coats (Fig. 4, B, B′, and E). This polarized transport relied on a system of motors and adaptors that would later evolve into the IFT system (Fig. 4, C and C′). The combination between the sensory and motility functions on the proto-cilium would culminate in the evolution of an organelle similar to those found in extant organisms (Mitchell, 2007; Satir et al., 2008).

Ample evidence supports the idea that sensory functions of the cilium evolved from polarized trafficking. The IFT system is likely to be a specialization of a more general transport system already present in the LECA: (a) coat proteins and small GTPases, typically found on the cytosolic face of organelles, have counterparts in the IFT machinery; (b) kinesins and dyneins involved in IFT have cytoplasmic counterparts involved in polarized transport (Wickstead and Gull, 2006, 2007; Wickstead et al., 2010); (c) some IFT components have nonciliary functions such as polarized secretion in nonciliated lymphoid cells and in exocytosis at the synapse of nonciliated secondary neurons in the retina (Finetti et al., 2009; Sedmak and Wolfrum, 2010); (d) the kinesin-II family of IFT motors also functions in polarized cells including intracellular melanosome transport and mRNA and MT stabilization (Wickstead and Gull, 2006; Jaulin and Kreitzer, 2010; Wickstead et al., 2010); and (e) several IFT proteins localize to spindle poles in mitosis, suggesting cytoplasmic roles in microtubule organization and spindle positioning (Fig. 3; Fig. 4, B, B′, C, C′, and E; Wickstead and Gull, 2006; Jaulin and Kreitzer, 2010; Wickstead et al., 2010; Delaval et al., 2011). Finally, cilia are specialized compartments, with different protein and lipid composition from those of cytoplasm and nonciliary membrane (Nachury et al., 2010). Diffusion barriers are known to play important roles in the definition of cellular compartments. It is plausible that the early creation of a diffusion barrier, perhaps organized by septins, which our unpublished data suggest are ancestral molecules, led to the individualization of a ciliary membrane (Fig. 4, B, C, C′, and E).

Less clear is whether the sensory patch predates the polarization of microtubules into a proto-cilium, whether the targeting of sensory and signaling molecules was a later adaptation, or even if they formed concomitantly. As the sensory and motility functions of the cilium appear to be widespread, it is not possible at this point to order their emergence in evolution.

Variations on the ancestral 9+2 axoneme configuration are observed among eukaryotes. Interestingly, these alterations correlate with changes in beat waveforms or loss of motility (Fig. 2). For instance, certain plants, diatoms, insects, and gregarines possess motile axonemes that lack a central pair, outer or inner dynein arms, or show deviations from the ninefold symmetry (Fig. 2; Manton et al., 1970; Heath and Darley, 1972; Baccetti et al., 1973; Schrevel and Besse, 1975; Prensier et al., 1980; Woolley, 1997; Okada et al., 2005; Riparbelli et al., 2009; Dallai et al., 2010). Although 9+2 axonemes use planar or three-dimensional beat waveforms, axonemes lacking the central pair typically have a simpler helical beat (Fig. 2; Holwill, 1966; Leadbeater and Dodge, 1967; Prensier et al., 1980; Goldstein and Schrével, 1982; Gibbons et al., 1983; Werner and Simmons, 2008). Nonmotile sensory cilia can show a variety of axoneme structures, as it is in the extreme case of nonmotile Caenorhabditis eleganscilia (Inglis et al., 2007). This observation suggests that the main constraint limiting the variation in axoneme architecture is their motility (Mitchell, 2007; Satir et al., 2008).

Evolutionary constraints applied to the axoneme structure are likely to be extended to the CBBs. For example, C. elegans, which does not have motile cilia, contains a small MT doublet “basal body” at the base of sensory cilia, whereas its embryonic centrioles, which are not associated with cilia, only have singlet MTs (Fig. 2). Similarly, Drosophila embryonic centrioles, which are not associated with cilia, are composed of MT doublets, whereas during motile flagella formation in spermatogenesis, CBBs that nucleate motile cilia are composed of MT triplets (Fig. 2; Callaini et al., 1997). In the extreme case, many species have completely lost their CBBs and cilia, such as Angiosperms, some fungi, and certain amoebas (Fig. 2).

The pathway of CBB biogenesis varies among eukaryotes being performed either in proximity to an already existing “parental”structure or de novo (Fig. 1 B). Canonical and de novo CBB assembly can also coexist in the same organism, such as in parthenogenic insects (Riparbelli and Callaini, 2003; Ferree et al., 2006). Because in animals and Chlamydomonas experimentally induced de novo biogenesis lacks control on the place, time, and number of CBBs assembled, it has been suggested that the presence of a parental structure provides a scaffold for the regulation of new centriole assembly (Marshall et al., 2001; Rodrigues-Martins et al., 2007b). It is thus striking how organisms such as the amoeboflagellate Naegleriaand sperm cells of Plasmodium, ferns, or diatoms that form CBBs de novo show impressive number, spatial, and time control (Fig. 2; Dingle and Fulton, 1966; Mizukami and Gall, 1966; Manton et al., 1970; Heath and Darley, 1972; Fritz-Laylin et al., 2010a). Perhaps in all organisms regulation of the location and numbers is achieved by controlled availability and localization of assembly components, which is often but not always specified by a parental CBB.

Several unicellular organisms assemble new cilia/flagella apparatuses during cell division without disassembling the old cilia/flagella and with no disruption of the ability to move. This is the case for ciliates, Trypanosoma, and Giardia, among others (Allen, 1969; Dute and Kung, 1978; Iftode and Fleury-Aubusson, 2003; Nohynková et al., 2006; Elias et al., 2007; Lacomble et al., 2010). Disassembly of cilia/flagella in mitosis, followed by reassembly in the next cycle is seen in unicellular and multicellular organisms, such as Chlamydomonas and human cells (Johnson and Porter, 1968; Cavalier-Smith, 1974; Rieder et al., 1979). It has been suggested that Chlamydomonas reabsorb their flagella due to the presence of a cell wall, which inhibits the migration of the basal bodies into opposite poles (Parker et al., 2010). In human cycling ciliated cells, where cilia disassembly is also observed, it seems plausible that this mechanism requires regulatory coordination with the cell cycle, though currently little is known about the mechanisms regulating this process.

The animal CBB has a dual function as an axonemal nucleator and as part of the cytoplasmic MTOC. In animals, the centriole, as part of the centrosome, has important functions in embryonic development, asymmetric divisions, and male meiosis (Rodrigues-Martins et al., 2008; Debec et al., 2010). A centrosome composed of centrioles and PCM is primarily found in animals and Fungi (Opisthokonts); however, other examples are found outside of this taxon, such as in brown algae and Plasmodiophorids (plant parasites belonging to the Cercozoa group; Fig. 2and Table S1). Although all species that present CBBs also present cilia, a weaker correlation is seen between the presence of the CBB and CBB-containing centrosomes. This observation suggests that the ancestral CBB function would be mainly to template/support the axoneme.

During cell division, the spindle ensures equal separation of the sister chromatids into the two daughter cells. The MTOCs that contribute to spindle MT nucleation and organization are highly variable among eukaryotes. The localization of CBBs at or close to the poles of the spindle is often achieved through the interaction of MTs nucleated by the CBB and the spindle itself. It has been suggested that this localization is a strategy to ensure equal inheritance of these structures so that both daughter cells can form cilia (Pickett-Heaps, 1971, 1974). However, even if this was originally the case, the CBB has important functions in cell division, suggesting possible cooption of the structure to actively participate in the process and coordinate it with other cell functions.

There is a core mechanism of CBB assembly conserved among eukaryotes determined by an evolutionarily cohesive and ancestral gene module (UNIMOD; Fig. 1 Band Fig. 3). The use of that module extends to different tissues within the same organism, as we and others have shown that canonical, de novo and assembly in multiciliated cells use the same molecular pathway to assemble the centriole structure (Rodrigues-Martins et al., 2007a; Vladar and Stearns, 2007; Kuriyama, 2009). Perhaps one of the most striking features of the evolution of CBBs is the almost perfect correlation between the absence of the structure and absence of that ancestral gene module that is strictly required to assemble the structure (Figs. 2and 3). Exceptions to this rule may provide interesting examples of eukaryotes that are in the process of losing the structure. That is the case of the smallest eukaryotic cell, the algae Ostreococcus, which does not seem to have a CBB but encodes some of its components in the genome (Figs. 2and 3; Henderson et al., 2007; Merchant et al., 2007; Wickstead and Gull, 2007; Keller et al., 2009; Carvalho-Santos et al., 2010; Hodges et al., 2010).

Despite the conservation of structural components, proteins that contribute additional regulatory steps, upstream or downstream of CBB assembly, appear to have been added in a stepwise, taxon-specific manner throughout evolution (Carvalho-Santos et al., 2010; Hodges et al., 2010). For example, SPD2/CEP192 is a centrosomal protein involved in PCM recruitment, which is only present in Amoebozoa and Animals (Kemp et al., 2004; Pelletier et al., 2004; Dix and Raff, 2007; Giansanti et al., 2008; Zhu et al., 2008; Carvalho-Santos et al., 2010; Hodges et al., 2010). SPD2/CEP192 is required for the duplication of the naked PCM-less sperm centriole in animals, suggesting a specific role at this developmental stage (Kemp et al., 2004; Pelletier et al., 2004; Dix and Raff, 2007). CP110 and CEP97, two centriolar proteins that have been proposed to coordinate ciliogenesis with the cell cycle, are confined to the Metazoans (Fig. 3; Carvalho-Santos et al., 2010; Hodges et al., 2010). Thus, these proteins might have been added to the pathway of CBB/cilium biogenesis as additional mechanisms to regulate this process through the cell cycle.

Additionally, gene duplication plays an important role in the evolution of CBB components and, in the case of BLD10/CEP135, this is a source of tissue specificity in CBB and flagella biogenesis (Carvalho-Santos et al., 2010). Protein divergence among orthologues may also contribute to species-specific adaptations in CBB assembly (Carvalho-Santos et al., 2010).

Comparative genomics, powered by the rapid increase in the availability of complete genome sequences from all branches of the eukaryotic tree, is revealing the nature of our eukaryotic ancestor and its evolution. As we sequence the genomes of more diverse eukaryotes, we also predict an increasingly complex LECA (Figs. 3 and 4; Fritz-Laylin et al., 2010b; Koonin, 2010). Because we have not yet found intermediate structures, we can only speculate how CBBs and cilia could have emerged from simpler, preexisting components. We hypothesize that the creation of diffusion barriers coupled with gene duplication and divergence may have provided a mechanistic driving force for the emergence of eukaryotic subcellular structures. Diversification of these structures and their function in different eukaryotes and different tissues likely proceeded through stepwise addition, duplication, and divergence of molecular components.

CBBs and derived structures are unequivocally identified by electron microscopy, in particular their characteristic ninefold symmetry, which has permitted their identification in different branches of the eukaryotic tree. These structures were lost in some Fungi, Amoebozoa, Alveolates, and Angiosperms, concomitant with loss of genes that are exclusively needed for CBB architecture such as SAS6, SAS4/CPAP, and BLD10/CEP135 (Figs. 2 and 3). This coarse correlation has been widely used for predicting new components of these structures (Avidor-Reiss et al., 2004; Li et al., 2004). As phylogenetic patterns for structural parts of CBBs and derived structures emerge, it will be important to investigate whether they correlate with certain molecular and morphological signatures (Fig. 2). For example, it is of key importance to piece together the evolution of the transition zone, the axonemal region where the transformation from basal body to axoneme occurs, because its components are often involved in many human diseases. This type of analysis should help us to identify new molecular components of this structure as well as having a better understanding of their mechanism of action.

The identification of pairs of closely related species where these structures were maintained in one organism and lost from the other (e.g., Thalassiosira vs. Phaeodactylum [flagellated vs. nonflagellated diatoms], Chlamydomonas vs. Ostreococcus [flagellated vs. nonflagellated green algae], and Plasmodium vs. Theileria [flagellated vs. nonflagellated apicomplexa]) may provide additional insights into how cellular structures and genes are lost during evolution. Finally, in the future it will also be important to understand how different cellular mechanisms have coevolved, in particular complexes determining cell polarity, protein trafficking, and the MT cytoskeleton. The centriole and its derived structures are thus an ideal paradigm to study the mechanisms involved in the evolution of the eukaryotic cell.

This paper contains a supplemental Table and corresponding references, available.

We would like to thank two anonymous reviewers, Lillian Fritz-Laylin, Benjamin Engel, Damien Devos, Max Nachury, and Yves Barral for discussions and critical reading of the manuscript. We would also like to acknowledge the authors of the EM images we used in this review, Ivan Vorobjev, Jennifer Sturgess, and Lotte B. Pedersen, as well as Bernhard Schermer and Thomas Benzing. We would like to thank Michel Bornens, Wallace Marshall, Bill Wickstead, and Keith Gull for discussions on this topic.

We are grateful to grants from Fundação Calouste Gulbenkian, Fundação para a Ciência e Tecnologia (FCT, POCI2010, PTDC/BIA-BCM/73195/2006, PTDC/BIA-BCM/105602/2008), and an EMBO Installation grant to M. Bettencourt-Dias. Z. Carvalho-Santos is recipient of a fellowship from FCT.

The authors declare no competing financial interests.