Antigenic variation of the parasite Trypanosoma brucei operates by monoallelic expression of a variant surface glycoprotein (VSG) from a collection of multiple telomeric expression sites (ESs). Each of these ESs harbours a long polycistronic transcription unit containing several expression site-associated genes (ESAGs). ESAG4 copies encode bloodstream stage-specific adenylyl cyclases (AC) and belong to a larger gene family of around 80 members, the majority of which, termed genes related to ESAG4 (GRESAG4s), are not encoded in ESs and are expressed constitutively in the life cycle. Here we report that ablation of ESAG4 from the active ES did not affect parasite growth, neither in culture nor upon rodent infection, and did not significantly change total AC activity. In contrast, inducible RNAi-mediated knock-down of an AC subfamily that includes ESAG4 and two ESAG4-like GRESAG4 (ESAG4L) genes, decreased total AC activity and induced a lethal phenotype linked to impaired cytokinesis. In the Δesag4 line compensatory upregulation of apparently functionally redundant ESAG4L genes was observed, suggesting that the ESAG4/ESAG4L-subfamily ACs are involved in the control of cell division. How deregulated adenylyl cyclases or cAMP might impair cytokinesis is discussed.
African trypanosomes are pathogenic protozoans causing human sleeping sickness and related diseases of livestock. These extracellular parasites multiply in the bloodstream of the vertebrate host, and escape complement-mediated lysis by a sophisticated system of variation of their antigenic surface coat, which is composed of homodimers of a variant surface glycoprotein (VSG). Bloodstream forms (BFs) of Trypanosoma brucei express only one VSG gene at a time from a collection of more than 1500 genes and pseudogenes, and this expression occurs in one of multiple (∼ 20) telomeric VSG expression sites (ESs) (Barry et al., 2005; Pays, 2006; Taylor and Rudenko, 2006). The different ESs all contain a polycistronic transcription unit with up to 12 different expression site-associated genes (ESAGs) in addition to the VSG (Pays et al., 2001). Like the VSG, these genes are only expressed from the active ES in BFs. A particular subfamily of T. brucei adenylyl cyclases (ACs) is encoded by one of the ESAGs, namely ESAG4, and is therefore only expressed in BFs (Pays et al., 1989). In contrast to ESAG4, other AC family members termed GRESAG4s (genes related to ESAG4) are constitutively transcribed throughout the life cycle of the parasite (Alexandre et al., 1990; 1996), and due to their chromosome-internal position are not subject to the allelic exclusion of the ES. Therefore, multiple GRESAG4s are simultaneously present in the plasma membrane of BF trypanosomes (Bridges et al., 2008).
The presence of key enzymes of the cyclic adenosine 3′, 5′-monophosphate (cAMP) metabolism in T. brucei suggests that cAMP signalling operates in this organism. ESAG4 was the first T. brucei AC whose function was confirmed by functional complementation of thermosensitive yeast mutants deficient in AC activity (cyr1-2) (Ross et al., 1991; Paindavoine et al., 1992). The overall structure predicted for ESAG4 is that of a surface receptor with a large variable extracellular domain of around 900 amino acids, separated from the highly conserved cytoplasmic catalytic domain (around 350 amino acids) by a single transmembrane helix (classified as a class II AC; Tang and Gilman, 1992). It was speculated that this extracellular domain could transduce signals from the host (Paindavoine et al., 1992). Immunogold electron microscopy (EM) detection of ESAG4 revealed a preferential localization on the flagellum (Paindavoine et al., 1992). The structure of ESAG4 appears to be shared by the other members of the family, and resembles that of receptor guanylate cyclases, except for the absence of a pseudokinase homology domain (reviewed in Seebeck et al., 2004). Apart from ACs, other enzymes involved in cAMP metabolism, such as phosphodiesterases (PDEs), have also been identified in T. brucei (Zoraghi et al., 2001; Rascon et al., 2002; Zoraghi and Seebeck, 2002; Kunz et al., 2004; Oberholzer et al., 2007). However, classical components of the cAMP signalling pathways, such as G protein-coupled receptors and heterotrimeric G proteins, are missing (Parsons and Ruben, 2000; El-Sayed et al., 2005). Moreover, apart from the GAF-A domain of PDEB2 (Laxman et al., 2005), no cAMP downstream effector candidate has been found so far. The identified protein kinase A (PKA)-like kinases (Parsons and Ruben, 2000; Kramer et al., 2007) are not regulated by cAMP in vivo (S. Bachmaier, S. Kramer and M. Boshart, unpublished). This is likely due to the presence of several point mutations in each of the two cyclic nucleotide monophosphate (cNMP)-binding domains of the single regulatory subunit PKAR (Shalaby et al., 2001).
It was hypothesized that the battery of ESAGs constitutes a tool for optimizing the adaptation of bloodstream-form parasites to their host, as suggested in the case of ESAG7/6, which encode a heterodimeric receptor for transferrin (Tf), a vital growth factor for the parasite (Bitter et al., 1998). The slight amino acid differences between ESAG7/6 copies expressed from different ESs lead to drastic changes of affinity for Tf (Salmon et al., 1997), possibly ensuring optimal Tf uptake in different host species. Although this hypothesis is still a matter of debate (Salmon et al., 2005; Steverding, 2006), the existence of a collection of slightly different ESAG4 copies in the various ESs might allow the parasite to regulate growth in response to different environmental conditions (Borst and Fairlamb, 1998; Pays et al., 2001).
The developmental life cycle of the parasite requires drastic morphological and biochemical changes and cycling between multiplicative and non-proliferative stages both in the mammal and in the tsetse fly vector (Fenn and Matthews, 2007). In the mammalian bloodstream, T. brucei proliferate as long slender forms, and then differentiate into non-dividing short stumpy forms that are pre-adapted for survival in the insect. The latter forms cannot undergo antigenic variation (Amiguet-Vercher et al., 2004) and are killed by the host immune system. This differentiation is independent of the antibody response (Seed and Black, 1997), but appears to depend on quorum sensing mediated by a low-molecular-weight component (stumpy induction factor, SIF) secreted by the trypanosomes themselves (Vassella et al., 1997). During this transition a peak of intracellular cAMP is produced (Mancini and Patton, 1981; Vassella et al., 1997). A cell-permeable analogue of cAMP (8-pCPTcAMP) has also been shown to induce the differentiation process (Vassella et al., 1997). However, only hydrolysable cAMP analogues appear to prevent cell proliferation, suggesting that breakdown products of cAMP, rather than bona fide cyclic nucleotides, are responsible for the antiproliferative properties transforming slenders to stumpy-like forms (Laxman et al., 2006). In the tsetse fly, stumpy forms differentiate into procyclic forms (PFs) and resume proliferation for colonization of the insect vector. In vitro, stumpy forms can also efficiently differentiate into PFs and this process is induced by cold shock that sensitizes trypanosomes to the differentiation inducers cis-aconitate and citrate (Engstler and Boshart, 2004). During this stumpy to PF differentiation, two distinct peaks of AC activity have been observed, the first occurring 6–10 h after the triggering of differentiation and the second between 20 and 40 h, when the cells that emerge from the first division start multiplying (Rolin et al., 1993). One of the immediate consequences of this differentiation is the modification of the major parasite surface proteins: the replacement of the VSG coat by procyclin, coinciding with downregulation of ESAG4. Although the release of VSG and activation of ACs occur simultaneously, these processes appear to be independent, both being direct consequences of cellular stress (Rolin et al., 1996). In spite of these correlations between the activation of ACs and cellular differentiation in T. brucei, the role of cAMP as a second messenger remains unclear, as does the significance of the presence of multiple AC isoforms. In this report we use an ESAG4 knockout (KO) cell line to show that this AC is neither essential for viability and growth nor important for differentiation. In contrast, knock-down of the ESAG4 subfamily (ESAG4+ESAG4L) induced a severe impairment of cytokinesis and lethality.
The repertoire of ACs in the T. brucei genome
The T. b. brucei genome includes a collection of 15–20 BF-specific ESAG4 AC genes (Hertz-Fowler et al., 2008). A large family of non-telomeric ESAG4-related genes (GRESAG4s) encoding around 60 different translatable isoforms (Figs 1 and S1) was identified by in silico analysis of the TREU927 reference genome (Berriman et al., 2005). Most of these genes are clustered in the genome, frequently in tandem arrays. Among the previously described genes, GRESAG4.1 and GRESAG4.4 (Alexandre et al., 1996; Naula et al., 2001) represent two major subfamilies with 13 GRESAG4.1 members on chromosome VI and six GRESAG4.4 members on chromosome IV, respectively, interspersed by pseudogenes. The tandem arrangement suggests that the two major subfamilies of ACs result from recent gene duplications. This is also supported by the phylogenetic tree (Figs 1 and S1). The previously described genes GRESAG4.2 and GRESAG4.3 (Alexandre et al., 1996) form a small subfamily spread along chromosomes VII and X respectively (Figs 1 and S1). More GRESAG4.4 genes reside on chromosomes VII and VIII. The subfamily of ES-associated ESAG4 forms a clade with particularly short branches. The nucleotide sequence identity among the ES-associated ESAG4s ranges from 94% to 99%. This is probably due to the homogenizing effect of frequent reciprocal recombination among the telomeric ES. Two non-telomeric GRESAG4 genes (Tb927.10.16190, Tb927.11.17040) belong to the same clade formed by telomeric ESAG4s. The encoded genes share 81% nucleotide sequence identity (78% amino acid identity) with one of the two ESAG4 copies from the Lister 427 VSG 221 ES (BES1/TAR40.13; geneDB H25N7.26, arrow in Fig. 1), suggesting some functional redundancy between ESAG4 and the two related GRESAG4 ACs, termed ESAG4L (ESAG4-like). In addition to the GRESAG4 genes shown in Fig. 1, GRESAG4 pseudogenes are distributed among most of the megabase-sized chromosomes, in particular chromosomes I, II and III that are devoid of functional GRESAG4s. The sequence alignments (Fig. S2A) of AC C-termini, which are highly divergent between isoforms, revealed a small region conserved in 11 of 15 ESAG4s. This peptide (highlighted in Fig. S2A) was used to raise ESAG4-specific antibodies. Similarly, antibodies were generated against a peptide characteristic for four members of the GRESAG4.4 AC subfamily (highlighted in Fig. S2B). Western blot analysis revealed that the GRESAG4.4 isoforms are expressed in both life cycle stages of the parasite, whereas ESAG4 is exclusively expressed in BFs (Fig. 2E), in accordance with previous observations (Alexandre et al., 1990; Paindavoine et al., 1992).
ESAG4 expression is not essential for growth in culture and parasite survival in mice
In order to assess the function of bona fide ESAG4 in a pleomorphic line of T. brucei, we first generated a knockout of ESAG4 in the active VSG expression site of AnTat 1.1E (abbreviated as ESAG4 KO in the following text). ESAG4 was replaced with a gene encoding resistance to neomycin (NEO) (Fig. 2A). To compensate for low targeting efficiency, the selection of pleomorphic transformants was performed in immunocompromised mice treated with the neomycin analogue G418 (Uzureau et al., 2007). After two rounds of selection in mice, two clones were subjected to further analysis. Probes covering the entire ESAG4 ORF or only the 5′ UTR of this gene (Fig. 2A) identified a specific band in Southern blots of AnTat 1.1E genomic DNA not present in the ESAG4 KO (clone G1) genomic DNA (filled arrowheads in Fig. 2B). Hybridization with a probe for NEO confirmed the integration of the selection marker into the ESAG4 locus (hollow arrowheads in Fig. 2B). Northern blot analysis using high stringency hybridization showed expression of the NEO gene (data not shown), whereas a probe spanning the entire ESAG4 ORF revealed a significant decrease of ESAG4 mRNA in the KO parasites (Fig. 2C). The residual amount of mRNA could be attributed to the cross-detection of related AC transcripts, since a similar signal, albeit weaker, was also observed with RNA from PFs where ESAG4 is not transcribed (Fig. 2C, see also Pays et al., 1989). Further evidence was provided by RT-PCR analysis (Fig. 2D) using primers amplifying a conserved 234 bp fragment at the 3′ end of the ESAG4 ORF (see alignment in Fig. S2C). Whereas a 0.7 kb tubulin internal control product was amplified similarly from both wild-type (WT) and KO cells, the ESAG4-specific PCR product was only present in WT parasites (arrowheads in Fig. 2D). Finally, Western blot analysis using anti-ESAG4 serum showed apparent absence of ESAG4 in the KO, also confirming the antibody specificity (Fig. 2F). It follows that a targeted ESAG4 gene deletion was generated and maintained in the active ES under selective pressure. ES switching was excluded since the AnTat1.1 VSG was still detected in the transformed parasites (Western not shown). A possible reactivation of ESAG4 expression from silent ES would be visible in Fig. 2D and also for most ES in Fig. 2F (see also Fig. S2). Therefore, the data suggest that the allelic exclusion of the expression site was fully operative on ESAG4.
In order to investigate the function of ESAG4 in BFs, we analysed the growth properties of the ESAG4 KO cell line. In vitro, cell growth in methylcellulose medium (Fig. 3A) was almost unaffected, with average parasite doubling times of 7.67 ± 0.53 h and 7.11 ± 0.39 h for KO and WT cells respectively (mean ± SE, n = 3). The growth of the AnTat 1.1E WT and ESAG4 KO lines was also compared during the course of infection in immunocompetent NMRI mice (Fig. 3B and C). Great care was applied to compare populations with similar selection histories and number of passages (see Experimental procedures). A hemizygous T. brucei AnTat 1.1E line with one allele of the kinesin TbKHC1 (Tb927.6.4390) replaced by NEO (Uzureau et al., 2007) was included as a control. Mice infected with ESAG4 KO or Δkhc1/KHC1 clones were treated with G418 during the first 3 days of each week of infection to prevent ES switching. Infection courses for ESAG4 KO and the controls showed a maximal first peak of parasitaemia around 1.2-2.5 108 cells ml−1 blood (Fig. 3B) and no difference was seen for up to 30 days of infection (Fig. 3C). T. b. rhodesiense exposed to normal human serum (NHS) invariably switches to an active ES, termed R-ES, that is devoid of several ESAGs including ESAG4. The R-ES encodes a specific ESAG, termed SRA, which confers parasite resistance to NHS (Xong et al., 1998). Northern blot analysis of several isolates of T. b. rhodesiense showed that ESAG4 and SRA mRNA expression is mutually exclusive (Fig. 2G). This provides further evidence that the presence of an ESAG4 copy in the active ES is not essential for growth or virulence in the host. In agreement, two monomorphic T. b. rhodesiense clones, ETat 1.2S and ETat 1.2R, either expressing ESAG4 or not, showed population doubling times of 8.3 ± 0.0 h and 8.4 ± 0.0 h, respectively, in HMI-9 medium supplemented with bovine serum (Salmon et al., 2005).
The course of infection in immunocompromised mice was also investigated (Fig. 3D). Any impairment of differentiation from the slender to the cell cycle-arrested stumpy bloodstream stage would have resulted in a higher first peak of parasitaemia, similar to the behaviour of monomorphic trypanosome strains. Figure 3D shows that pleomorphic ESAG4 KO cells and control AnTat 1.1E cells reached the same maximal density in immunocompromised mice. The expression kinetics of a specific stumpy stage marker, mitochondrial NADH diaphorase activity did not differ between ESAG4 KO and AnTat 1.1E cells (Fig. 3E). We conclude that ESAG4 is not required for cell density-dependent entry into the G0 phase of the cell cycle or for induction of slender to stumpy differentiation.
Members of the ESAG4+ESAG4L subfamily of cyclases are functionally redundant
The absence of a growth phenotype of ESAG4 KO may be due to selection for compensatory mechanisms during the expansion of the KO clones. A conditional knock-down experiment may thus have a different outcome (Helms et al., 2006). Selective RNAi-mediated repression of ES-associated ESAG4 is however not possible due to the closely related ESAG4L genes (see Fig. 1). Therefore, we targeted the entire subfamily including ESAG4 and the ESAG4L genes Tb927.10.16190 and Tb927.11.17040. We identified a suitable targeting region and ascertained the RNAi selectivity for this subfamily by processing the full repertoire of AC genes from the TREU927 and Lister 427 genomes by custom written RAiN software (M. Kador and M. Boshart, available as web service at http://boslinux.gi.biologie.uni-muenchen.de/rain/RNAi/index.html). This application identifies regions of non-identity, which are suitable for isoform-specific RNAi-mediated silencing within large gene families. By using a scanning window of 25 nt, the mean size of siRNAs in T. brucei (Djikeng et al., 2001), we identified a fragment for targeting of ESAG4 and ESAG4L but not any other AC family members. The RAiN reanalysis of the AC subfamilies using this fragment as query is provided in Fig. S2D. The fragment was amplified from T. brucei 427 221 (MiTat 1.2) DNA (BES1/TAR40.13, geneDB H25N7.26; in this strain ESAG4 is duplicated), subcloned into a p2T7 vector for inducible RNAi (LaCount et al., 2000) and introduced into the 427 13-90 monomorphic T. brucei cell line (Wirtz et al., 1999). Two independent clones (#1.2; #4.3) were analysed. Quantification of mRNA levels by qRT-PCR, after induction with tetracycline for 8 h, showed a fivefold and a twofold reduction of ESAG4 transcripts in the independent clones #4.3 (relative expression ratio rER of 0.22) and #1.2 (rER of 0.55) respectively. RT-PCR analysis was highly specific for ESAG4 mRNA as indicated by a rER of 0.00048 between PFs and BFs of strain AnTat 1.1. Northern blot hybridization confirmed a significant reduction of ESAG4 mRNA in both RNAi clones in an independent experiment after 8 h of induction (Fig. 4A). The ESAG4 protein disappeared almost completely from Western blots of induced samples after 72 h (Fig. 4B). Induced cells showed a severe growth phenotype and subsequent cell death (Fig. 4C). This result was independent of the active ES context, as inducible RNAi in T. brucei MiTat 1.4 resulted in an almost identical growth phenotype (Fig. S3). A hairpin construct (pHD615PAC.ESAG4) targeting a similar region of ESAG4 was transfected in that case. The very reproducible and strong RNAi phenotype contrasted with the apparent absence of a growth phenotype in the ESAG4 KO line. To exclude the possibility of cross-RNAi, we replaced one of the two tandem copies of ESAG4 in the 221 (MiTat 1.2) ES in situ with a gene version recoded to a non-homologous nucleotide sequence in the RNAi targeting region (RAiN reanalysis see Fig. S2D) followed by transfection of this line with the pT2T7.ESAG4A RNAi construct. Upon tetracycline induction, no growth phenotype was observed in three independent subclones (Fig. 4D). Furthermore, the functionality of the ESAG4+ESAG4L RNAi in two of the subclones with silent recoding of ESAG4 (#5.1 and #5B6) was verified by qRT-PCR analysis of both ESAG4L genes (Fig. S4). The degree of repression was similar in the recoded subclones and in the original RNAi line shown in Fig. 4E. In contrast, ESAG4 mRNA was not affected by RNAi in the recoded subclones. The complete rescue by silently recoded ESAG4 allows us to conclude that the ESAG4+ESAG4L subfamily is responsible for the observed RNAi phenotype. Therefore, functional redundancy among the members of the ESGAG4+ESAG4L subfamily or selection for compensatory changes rescuing the deficiency in the ESAG4 KO line were considered. The expression of ESAG4+ESAG4L and representative GRESAG4 genes was quantified by qRT-PCR. The data are presented as the relative expression ratio either between uninduced and induced RNAi or between WT and KO (Fig. 4E). For the GRESAG4.1 and 4.4 subfamilies, which are not targeted by our RNAi strategy, qRT-PCR primers specific for a subset of genes belonging to the respective subfamily were designed. The values were almost unchanged upon RNAi, as expected (rER ∼ 1). RNAi induction resulted in low rER for ESAG4 and the ESAG4L (Tb927.10.16190 and Tb927.11.17040), documenting the effective repression of the subfamily. When the ESAG4 KO cell line was compared with its WT control, a very low rER was obtained for ESAG4, as expected. Surprisingly high rER values indicated significant overexpression both of ESAG4L and to a lesser degree of GRESAG4.1 subfamily members. We concluded that targeted deletion of the expressed ESAG4 copy resulted in selection for significant overexpression of ESAG4L, most likely as a compensatory mechanism to restore a critical level of cyclase expression. Direct evidence for this interpretation was provided by total cyclase activity measurements (Table 1) using the swell dialysis method (Voorheis and Martin, 1980). After 4 h of RNAi induction, adenylyl cyclase (AC) activity was reduced two- to fourfold, compared with the uninduced control. In contrast, the total AC activity levels measured in the ESAG4 KO line and in the corresponding WT were very similar to each other.
Table 1. Adenylyl cyclase activity in ESAG4 KO and ESAG4+ESAG4L RNAi cell lines.
The ESAG4+ESAG4L subfamily is required for cell division of bloodstream forms
To investigate the observed growth phenotype, cell division was analysed by scoring the kinetoplast/nucleus (K/N) configurations by DAPI staining of tetracycline-induced or uninduced populations of ESAG4 RNAi clones. A strong increase of the fraction of 2K2N cells (between mitosis and cytokinesis) was found upon RNAi induction (maximum of 40% 2K2N cells after 6 h, Fig. 5A). The fraction of 1K1N cells (in G1 phase of cell cycle) dropped, whereas the fraction of 2K1N cells (S to early M phase) remained constant. In addition, some aberrant K/N configurations such as enlarged cells containing multiple nuclei and kinetoplasts were observed at later time points. In BFs of T. brucei, these cells can accumulate when cytokinesis is impaired due to rereplication in the absence of a cell cycle checkpoint restricting this (Hammarton et al., 2003). The data in Fig. 5A thus suggest that the cells were severely delayed or impaired in cytokinesis. This phenotype was fully rescued in ESAG4 RNAi clone 5.1 harbouring a silently recoded ESAG4 copy in the ES (Fig. 5C). More detailed analyses of the induced 2K2N fraction scored the accumulation of cells stalled in specific subphases of cell division (Fig. 5B and D). The fraction of cells having just finished mitosis (no visible cleavage furrow) decreased from 82 ± 2.5% to 39 ± 1.67% (n = 4) of all 2K2N cells. At the same time, we observed an increase from 8.4 ± 1.05% to 34 ± 0.42% of cells classified as furrowing (nomenclature according to Hammarton et al., 2007b). Cells in the process of abscission rose from 8 ± 1.37% to 21 ± 0.30%. We attempted to correlate the preferential RNAi vulnerability of cleavage furrow progression with expression of the cyclases along the cell cycle of WT cells. A proliferating bloodstream-form population was sorted according to DNA content by high speed FACS (see supplementary methods) to obtain G1 and G2/M populations (Fig. S5A). Expression of ESAG4+ESAG4L was then quantified in G1 and G2/M by qRT-PCR (Fig. S5B) and by Western blotting (Fig. S5C). ESAG4 mRNA expression per cell increased 2.25-fold and ESAG protein expression per cell increased 2.45-fold from G1 to G2/M, as expected from the gene dosage and cell volume in G2/M. A similar mRNA increase was noticed for the ESAG4L. Thus, no cell cycle-dependent expression regulation of the ESAG4+ESAG4L subfamily was observed. This is consistent with a recent RNA-seq study that showed no cell cycle regulation of any of the GRESAG4 genes (Archer et al., 2011). As an alternative to specific regulation, a critical absolute level of expression of ESAG4+ESAG4L throughout the cell cycle might be important for efficient cytokinesis and cleavage furrow progression in bloodstream trypanosomes. In agreement with this interpretation, all our attempts to stably introduce the silently recoded ESAG4 and genuine ESAG4 in high level constitutive and inducible vector systems have failed (Table S1 and data not shown), suggesting lethality of even moderate overexpression. Only the strictly regulated pLew100v5b1d-BSD.ESAG4rec vector gave transformants, and co-induction of RNAi and silently recoded ESAG4 resulted in a partial rescue of the RNAi phenotype (Table S1). Full rescue was only achieved by in situ replacement of a silently recoded ESAG4 copy (see Figs 4D and 5C). Together, the data suggest that total expression levels of ESAG4+ESAG4L are critical for cell division and viability.
The ESAG4+ESAG4L subfamily is developmentally regulated
The cell division phenotype prompted us to determine the expression of ESAG4 and ESAG4L in the cell cycle-arrested stumpy stage of the parasite that is locked in the G0 phase. The qRT-PCR results showed that ESAG4 mRNA expression in stumpy forms of strain AnTat1.1 amounts to only 24% of that in slender forms (Fig. 6A). ESAG4 protein expression was reduced to 7.6% in stumpy forms as determined by Western blot analysis (Fig. 6B). In a stumpy-like population of the monomorphic strain MiTat 1.2 at maximal culture density, ESAG4 protein expression was reduced to 44% compared with the proliferating slender population of that T. brucei strain (Fig. 6B). The specific downregulation of ESAG4 in stumpy forms corresponds to the previously reported downregulation of VSG mRNA and changes in the active ES in stumpy forms (Amiguet-Vercher et al., 2004). However, the same degree of downregulation was found for the two chromosome-internal ESAG4L. The mRNA of Tb927.10.16190 and Tb927.11.17040 was reduced in stumpy forms relative to slender forms, to 22% and 18% respectively. Also, a significant downregulation of mRNA of GRESAG4.1 (33%) and GRESAG4.4 (34%) subfamily members was noted (Fig. 6A). Clearly, silencing of the ES in stumpy forms is not compensated for by an upregulation of ESAG4L, as is seen in the ESAG4 KO line (Fig. 4E). This suggests that AC activity encoded by this subfamily is less important for the quiescent stumpy stage that is most responsive to differentiation signals (Engstler and Boshart, 2004; Fenn and Matthews, 2007). Therefore, it was not surprising that inducible RNAi repression of ESAG4+ESAG4L in the stumpy-like parasites of strain MiTat 1.4 did not change the kinetics of cis-aconitate-triggered procyclin expression, an early differentiation marker (data not shown). High level AC expression is correlated with the proliferative slender stage, and specifically the expression of the ESAG4+ESAG4L subfamily is essential for efficient cell division.
Almost 20 years after the discovery that ESAG4 encodes an AC and that numerous genes encoding related AC isoforms (GRESAG4s) are transcribed in both PF and BF trypanosomes (Pays et al., 1989; Alexandre et al., 1990), the signalling function of ACs in T. brucei remains enigmatic. Here, reverse genetic analysis of ACs in T. brucei showed that a critical level of expression of bloodstream stage-specific ESAG4 and ESAG4L is essential for efficient cytokinesis.
Why does an AC gene reside in most VSG ESs?
The absence of any discernible phenotype upon genetic ablation of ESAG4 from the active ES shows that this particular AC isoform is neither essential for proliferation in culture or in the mammalian host nor essential for slender to stumpy differentiation and, most importantly, that the ES-specific expression of AC is not essential. This is consistent with recent genomic analyses of the ES repertoire of strain Lister 427 (Hertz-Fowler et al., 2008). Six out of the 14 sequenced ESs lack an ESAG4 gene or contain only a pseudogene. The active ES in NHS-resistant T. b. rhodesiense also lacks ESAG4 (Xong et al., 1998). The question remains why the presence of ESAG4 in the ES has provided sufficient selective advantage to explain the spreading of the AC family into this very particular genomic localization in the course of evolution. As initially proposed for the ESAG7/6-encoded Tf receptors, slightly different isoforms of ESAG4 in different ESs may enable adaptation to host-specific putative ligands of ESAG4 and thus expand the host range (Borst and Fairlamb, 1998; Pays et al., 2001). However, a clear correlation between the host range and either the number of ESs or the extent of ESAG genetic diversity in several trypanosome species is missing (Young et al., 2008). ESAG4 does not seem to be a target of neutralizing host antibodies (D.S. and E.P., unpubl. data) and so far there is no evidence for any extracellular molecules specifically activating trypanosome ACs. We suggest that the advantage for the parasite in harbouring AC genes in the ES is the high BF-specific transcription rate of this polycistronic unit. Our results show that maintenance of a critical level of expression of ESAG4+ESAG4L is required for successful cell division. Targeted deletion of ESAG4 in the active ES was compensated by adaptive overexpression of the several non-telomeric ACs. RNAi-mediated knock-down of ESAG4+ESAG4L resulted in a sixfold reduction of total AC activity. Therefore, the contribution of the majority of constitutively expressed GRESAG4 isoforms to the total adenylyl cyclase activity in the bloodstream stage of WT cells seems below average, in spite of the high number of genes. Localization of ESAG4 in the ES may therefore confer a selective advantage for parasite fitness. For example, the quiescent metacyclic trypanosomes from tsetse saliva use monocistronic ESs for VSG expression (Graham et al., 1993; Alarcon et al., 1994). Switching to a polycistronic ES containing an ESAG4 gene may boost ESAG4 expression just upon entry into the proliferative phase, when a critical level of expression of ESAG4+ESAG4L is required.
GRESAG gene family expansion
Trypanosoma brucei ACs are encoded by a large multigene family of more than 80 members including at least five pseudogenes, and many of these isoforms are present at the cell surface simultaneously (Bridges et al., 2008). This gene family expansion appears to be specific to salivarian trypanosomes. Whereas the genomes of T. b. gambiense, T. vivax and T. congolense have similar numbers of AC genes, the genomes of intracellular trypanosomatids like Leishmania and Trypanosoma cruzi contain less than 11 AC genes (see TriTrypDB, version 4.0). Thus, the gene family expansion is correlated with the extracellular lifestyle and continuous exposure of the parasites to the defence systems of the mammalian host. This suggests that the host immune system may be the evolutionary force driving AC gene family expansion. We can envisage two reasons for gene amplification of ACs: (i) functional diversification and (ii) increase of gene dosage. It has been speculated that trypanosomal ACs, due to their predicted structural topology, may function as monomeric surface receptors/transducers for host-derived and environmental signals (Paindavoine et al., 1992; Alexandre et al., 1996; Seebeck et al., 2004). In this scenario, AC diversity would compensate the absence of canonical G protein-coupled receptors and heterotrimeric G proteins from trypanosomatid genomes. As mentioned before, no evidence for ligands is available and it is difficult to imagine that diversity of signalling specificities has only evolved in salivarian trypanosomes. We therefore favour the hypothesis that evolution has selected gene duplications and amplifications of the GRESAG4 ancestors, primarily to achieve high total AC expression. Later and for the same reason, the highly transcribed primordial VSG ES may have been invaded by one of the ancestral ESAG4L, giving rise to ESAG4. This scenario is fully compatible with the phylogenetic analysis of the AC family.
The ESAG4+ESAG4L subfamily and cytokinesis
In T. brucei the extreme cellular polarization, the singular organelles whose division is co-ordinated in the cell cycle and the tubulin exoskeleton render cytokinesis particularly vulnerable to defects in many processes. Due to the absence of the general ‘end of cytokinesis’ checkpoint (Hammarton et al., 2003), rereplication prior to completion of cytokinesis results in ‘monster’ cell (multiple nuclei, kinetoplasts and flagella) phenotypes and ultimately cell death. Therefore, distinguishing between direct and indirect effects on cytokinesis requires functional analysis at the earliest possible time points to determine the precise role of a protein in the process (reviewed by Hammarton et al., 2007b). RNAi repression of proteins thought to be directly involved in cytokinesis like MOB1, PLK and the NDR kinases PK50 and PK53 show an early increase of 2K2N cells (33%, 30%, 20%, 25%, respectively, after 8 h of induction in BF) and two- to fourfold more furrowing 2K2N cells (Hammarton et al., 2005; 2007a; Ma et al., 2010). VSG RNAi results in a complete precytokinesis block (55% of 2K2N, no furrowing cells after 8 h of induction), indicating a checkpoint triggered process (Sheader et al., 2005). Accordingly, a cytokinesis phenotype via perturbation of VSG synthesis has been found upon GPI8 RNAi, targeting GPI anchor biosynthesis (Lillico et al., 2003). Although direct comparability of these data may suffer from differences in RNAi efficiency, protein half-lives and microscopical scoring, the ESAG4 RNAi results (40% 2K2N and more than fourfold increase of furrowing cells 8 h post induction) can be placed between the complete initiation block of VSG RNAi and the furrow ingression phenotypes and would be compatible with a direct impact on cytokinesis. RNAi of a large number of proteins involved in flagellar motility and biogenesis showed strong indirect cytokinesis phenotypes (Broadhead et al., 2006; Ralston et al., 2006; Hammarton et al., 2007b). Microscopic analysis of ESAG4 RNAi cells has not even revealed a subtle motility phenotype, suggesting that such an indirect effect via motility is unlikely. We propose that ESAG4+ESAG4L cyclases might be involved in sensing of the VSG coat density and thereby control cytokinesis. This speculation is supported by a perfect correlation between the presence of a VSG coat and the number of cyclases or abundant AC expression, when comparing salivarian and non-salivarian trypanosomatids (Fig. 1) or proliferating developmental stages of T. brucei (Fig. 6). Sheader et al. (2005) suggested a cell cycle checkpoint that senses a compromised VSG coat or VSG synthesis/trafficking to the surface to adjust VSG availability and cellular surface growth and prevent coat dilution. This checkpoint mechanism may operate via the general but reversible translation initiation arrest seen upon VSG RNAi (Smith et al., 2009). ESAG4+ESAG4L cyclases are excellent candidates for a role as sensor(s) of a compromised VSG coat. They are among the limited number of transmembrane proteins distributed on the BF surface together with VSG, they have a large extracellular domain in demand of a function and they are signalling proteins whose AC activity is well known to be activated by membrane stress (Voorheis and Martin, 1982; Rolin et al., 1996; Nolan et al., 2000). In addition, overexpression of ESAG4 (Table S1 and data not shown) as well as RNAi-mediated depletion of the phosphodiesterases PDEB1/B2 (Oberholzer et al., 2007) or specific inhibition of PDEB1/B2 (de Koning et al., 2012) also resulted in lethality with a similar cytokinesis disturbance. Together this suggests a role of cAMP signalling in cytokinesis. We hypothesize that ESAG4+ESAG4L may be involved in a process coupling cell division to integrity of the VSG surface coat and therefore signal to a cell cycle checkpoint in the BF stage. We are currently testing this hypothesis by monitoring cAMP signalling during VSG RNAi.
Trypanosomes and transfections
Bloodstream forms of T. b. brucei were cultured at 37°C in modified HMI-9 medium (Vassella et al., 1997) supplemented with 10% (v/v) heat-inactivated fetal bovine serum. Alternatively, bloodstream stages were cultured in the above medium supplemented with 1.1% methylcellulose (Vassella et al., 2001). The following parasite strains were used: MiTat 1.2 and 1.4 from the T. brucei 427 Lister MiTat-serodeme (Cross, 1975) and AnTat 1.1E (Delauw et al., 1985). For cultures of the ESAG4 RNAi strain based on the cell line 427 13-90 (Wirtz et al., 1999), 5 µg ml−1 hygromycin, 2.5 µg ml−1 geneticin and 1 µg ml−1 phleomycin were added to the medium. For expression of a silently recoded ESAG4 version the following antibiotic selections were used: 0.1 µg ml−1 puromycin for ectopic, constitutive expression (pTSARibPAC.ESAG4rec); 5 µg ml−1 blasticidine for tet-inducible (pLew82BSD.ESAG4rec, pLew100v5b1d-BSD.ESAG4rec) or in situ silently recoded ESAG4 [in situ ESAG4rec (BSD)]. The MiTat 1.4 ESAG4 RNAi strain was cultured in HMI-9 medium containing 0.15 µg ml−1 puromycin and 1 µg ml−1 phleomycin. Electroporation and selection procedures were performed as described in Burkard et al. (2007). To obtain the AnTat 1.1E ESAG4 KO cell line, 2 × 108 parasites were transfected and injected intraperitoneally into two mice immediately after transfection. Selection of recombinant parasites was achieved by intraperitoneal injection of 30 mg kg−1 body weight day−1 G418 for 3 days, 24 h after parasite injection (Uzureau et al., 2007). Each mouse was immunosuppressed by injection of 35 mg kg−1 body weight cyclophosphamide on day 6 post injection. The emerging parasites were then passaged into fresh mice to undergo a second round of selection yielding completely resistant populations. After cloning on agarose plates, the trypanosomes were amplified twice in immunosuppressed NMRI mice treated with G418 to prevent ES switching. Stabilates were prepared from the second NMRI mouse passage. This stock was used for the infection experiment with sets of four mice. For the in vivo growth experiments of AnTat 1.1E WT, Δkhc1/KHC and ESAG4 KO trypanosome populations with comparable selection history on agarose plates and similar number of mouse passages were injected in order to eliminate non-specific changes of growth characteristics (Uzureau et al., 2007).
For microscopic analysis, 0.5 to 1 × 107 cells were fixed in 2% paraformaldehyde or 100% methanol. After washing with phosphate-buffered saline (PBS), cells were allowed to settle on glass slides. Cellular DNA was visualized with 4′,6-diamidino-2-phenylindole (DAPI; 30 µM). Image acquisition was performed with an Olympus IX81 microscope and a Panasonic WV-CL270 camera. The imaging software Olympus xcellence pro was used for image analysis. Cytochemical staining for NADH diaphorase activity was performed according to Vassella and Boshart (1996).
RNAi experiments were performed in T. brucei 427 13-90 (Wirtz et al., 1999) and in MiTat 1.4 449 [tet repressor construct pHD449 in the tubulin locus (pHD 449, Biebinger et al., 1997)]. Strain 427 13-90 was transfected with an RNAi construct containing an ESAG4 target fragment from BES1/TAR40.13, GeneDB accession H25N7.26, positions 1–526), inserted between the opposing T7 promoters of construct p2T7Ti for tetracycline-inducible dsRNA (LaCount et al., 2000). The target DNA fragment was amplified by PCR from the MiTat 1.2 BAC clone 40D16 (BACPAC Resource Center, Children's Hospital Oakland Research Institute; kindly provided by G. Rudenko). For hairpin RNAi, a similar target DNA fragment (GeneDB accession H25N7.26, positions 19–499) was amplified from genomic DNA in both orientations and cloned into the pHD615PAC vector.
An N-terminal fragment of 823 bp including the silently recoded RNAi target sequence (see Fig. S2D) of the ESAG4 copy BES1/TAR40.13 was chemically synthesized by GeneArt GmbH, now Invitrogen. This fragment replaced the respective original ESAG4 fragment in the target vectors pTSARibPAC, pLew100v5b1d-BSD, pLew82BSD and pTSARibBSD+ESAG45′UTR. For details see supplementary methods.
Amino acid sequences of 78 cyclase genes were aligned using clustal w 2.0 (Larkin et al., 2007) and edited manually using MacClade 4.08 (Maddison and Maddison, 2005). This amino acid sequence alignment was used as guide to align the corresponding nucleotide sequences. Ambiguously aligned regions (1170 characters out of 4475) were excluded prior to analysis. The included characters were analysed under the Maximum Parsimony (MP) criterion in PAUP* (Swofford, 2003). Starting trees were obtained via stepwise addition using 1000 replicates holding a single tree during each replicate. The tree-bisection-reconnection (TBR) was used with the ‘MulTrees’ option in effect. For the bootstrap analysis with 1000 bootstrap replicates the same parameters were applied but starting trees were obtained via stepwise addition using 20 replicates with the ‘MulTrees’ option not in effect. For the Bayesian analysis the General-Time-Reversible evolutionary model was used in combination with rate heterogeneity parameters allowing a proportion of invariable sites with remaining sites modelled according to gamma-distributed rate in MrBayes 3.1.2 (Ronquist and Huelsenbeck, 2003; Altekar et al., 2004). Independent runs were conducted in MrBayes for one million generations, sampling every 100 generations. The first 1500 sampled trees were discarded after inspection of the stationary curves, where likelihood scores were plotted against the number of generations in a spreadsheet programme.
Isolation of DNA and RNA from trypanosomes
Isolation of genomic DNA from T. brucei was performed using the method of Marmur (Marmur, 1961) or the NucleoSpin Tissue kit (Machery-Nagel, Düren) according to the manufacturer's instructions. Purification of total RNA was performed by guanidinium thiocyanate-phenol-chloroform extraction with either peqGOLD RNAPure reagent (Peqlab, Erlangen), Trizol (Invitrogen) or NucleoSpin RNA XS (Machery-Nagel, Düren) according to the manufacturers' instructions.
RT-PCR and qRT-PCR
Isolated RNA was treated with DNase prior to the reverse transcription reaction, using either TURBO DNase (Ambion, Texas, USA) or RQ1 RNase-free DNase (Promega, Mannheim) according to the manufacturers' instructions. The DNase was inactivated by addition of either 0.1 volume of DNase inactivation buffer for 2 min at RT or 1 µl of RQ1 DNase stop solution for 10 min at 65°C. Complementary DNA was synthesized with either the iScript cDNA synthesis kit for real-time RT-PCR (Bio-Rad) or the Transcriptor Reverse Transcriptase (Roche Applied Science) according to the manufacturers' instructions.
To monitor the absence of ESAG4 transcripts in the ESAG4 KO cell line, an ESAG4 fragment of 234 bp from the C-terminus (amino acids 1125–1203) was amplified with the oligonucleotides E4-3′spF (5′-CTTAACGATGGTGAAGATCAG-3′) and E4-3′spR (5′-CCGTACAACTTCCTGACAAC-3′) and a tubulin gene fragment of 789 bp was amplified as an internal control with the oligonucleotides TUBF (5′-AGGTACAATGGACTCCGTAC-3′) and TUBR (5′-CACGTTCAGCATCTGCTCAT-3′). Expand High Fidelity DNA Polymerase (Roche Applied Science) and the following cycling parameters: [2 min 95°C; 35 × (15 s 94°C, 30 s 54.3°C, 60 s 68°C); 7 min 68°C] were used.
Real-time PCR analysis was performed with either the iCycler iQ Real Time PCR Detection System (Bio-Rad, Munich) or the 7300 Real Time PCR System (Applied Biosystems) using the iQ SYBR Green Supermix (Bio-Rad, Munich) or the Platinum SYBR Green Quantitative Supermix-UDG kit (Invitrogen, Carlsbad, California, USA), respectively, according to the manufacturers' instructions. For quantification of ESAG4 mRNA levels, a 155 bp fragment was amplified with the oligonucleotides ESAG4.U3 (5′-ACCCCTCGGTGACCAAAGTG-3′) and ESAG4.L3 (5′-GAACCAATGTTTCAGCAGCGG-3′). One hundred nanograms of total RNA treated with DNase were used for reverse transcription and 10 ng of cDNA were amplified with the following cycling parameters: [5 min 95°C; 40 × (30 s 95°C, 30 s 60°C)]. Similar PCR conditions were used for quantification of Tb927.10.16190 and Tb927.11.17040 (ESAG4L) mRNAs resulting in fragments of 78 and 79 bp respectively, with the following oligonucleotides: ForTb10.61 (5′-ATTGCGGATAGCGGGATAATAAAC-3′), RevTb10.61 (5′-CGCCTGTGCTGACTGACC-3′) and ForTb11.01 (5′-GTAACAAACCTATCCACCAG-3′), RevTb11.01 (5′-GCATGACTGCCACTATCG-3′). Quantification of transcripts encoded by GRESAG4.1 and GRESAG4.4 subfamilies was performed under similar PCR conditions using oligonucleotides ForG4.1 (5′-TGTAGGTTTGATGTATGCGAAGG-3′) and RevG4.1 (5′-TCTCCACAGCGAACACTCC-3′) corresponding to NCBI GenBank Accession No. X52119 (amplicon length 113 bp, Alexandre et al., 1990), and ForG4.4 (5′-GCGTGAATTGTATGTTAGAGATGG-3′) and RevG4.4 (5′-TGGTAATAGATGCGTTGAAGCC-3′) corresponding to NCBI GenBank Accession No. AF228602 (amplicon length 188 bp; Naula et al., 2001). Primer pairs to GRESAG4.1 or GRESAG4.4 subfamilies were designed using Beacon Designer 7.0 in order to get GRESAG4 subfamily specificity. blastn 2.2.26+ searches of the T. brucei TREU 927 genome with the primer sequences confirmed that the only genes with sequence identity between both primers and target were Tb927.4.4440, Tb927.4.4410, Tb927.4.4430, Tb927.4.4450 (all in one clade, see Fig. 1) for GRESAG4.1 and Tb927.6.760, Tb927.6.790, Tb927.6.770, Tb927.6.780 (all in one clade, see Fig. 1) for GRESAG4.4 respectively. As a reference for normalization, a 212 bp fragment of the tubulin gene was amplified with the oligonucleotides Tub_up1_175 (5′-TATGTGCCCCGCTCCGTGCTG-3′) and Tub_low1_363 (5′-CAGTCACAGCTCTCCGCCTCCTTG-3′). For mRNA quantification during the cell cycle and the slender to stumpy differentiation, no internal markers with established constitutive expression are available. The normalization was done on cell numbers by spiking the samples with procyclic trypanosomes and using the stage-specific EP gene as a reference (Bucerius et al., 2011). For this purpose, 40 000 PCF cells were diluted in 1 × 106 BSF cells prior to RNA extraction. The EP cDNA was amplified using the oligonucleotides ForEP (5′-TCTGCTCGCTATTCTTCTG-3′) and RevEP (5′-CCTTGTCTTCTGGTCCTTC-3′).
All PCR reactions were done in duplicate and the average was used for data analysis. For the relative quantification of the Real Time PCR data, the PCR efficiencies were calculated with LinRegPCR 7.2 and the data were analysed with the 2−ΔΔCT method of Livak and Schmittgen (2001).
Production of anti-ESAG4 and anti-GRESAG4.4 antisera
Anti-ESAG4 and anti-GRESAG4.4 antisera were produced in rabbits against synthetic peptides coupled to a KLH carrier (Eurogentec) and corresponding to the following amino acid sequences: H2N-CVDGHCVTVQLIDLENDSATT-COOH and H2N-KYSPKHCNMCPLLPE-COOH. The anti-GRESAG4.4 antiserum lost sensitivity and specificity with time. It was used for Fig. 2E but not available for later experiments.
For quantitative Western blot analyses, lysates of 0.5 to 3 × 106 cells were separated on 10% polyacrylamide gels, blotted onto an Immobilon-FL PVDF membrane (Millipore) and probed with rabbit anti-ESAG4 serum, diluted 1:2500. Monoclonal anti-PFR-A/C antibody L13D6 (1:2000; Kohl et al., 1999) was used as an internal loading control. Alexa FluorTM 680-conjugated goat anti-rabbit (1:5000; Molecular Probes) and IRDye 800-conjugated goat anti-mouse IgG (H+L) antibodies (1:5000; LI-COR) were used as secondary antibodies. IR fluorescence signals were quantified with the OdysseyTM IR fluorescence scanning system (LI-COR). Signals of ESAG4 were normalized to the PFR-A/C loading control after automatic subtraction of the background values (Median Left/Right method) using the Odyssey software (LI-COR). Western blotting using enhanced chemiluminescence (ECL) for detection of HRP-conjugated anti-rabbit IgG (Promega) was performed on lysates of 5 × 106 cells separated on 10% polyacrylamide gels and blotted onto Nitrocellulose Hybond-C Extra membrane (Amersham).
Adenylyl cyclase assays
Trypanosomes (∼ 4 × 107) were induced for 4 h with tetracycline where indicated, and harvested by centrifugation, washed twice with PSG buffer (phosphate-buffered saline glucose containing 2.5 mM NaH2PO4·H2O, 47.5 mM Na2HPO4·2H2O, 36.5 mM NaCl, 1.5% glucose, pH 8.0), resuspended in 1 ml of PSG and counted. Adenylyl cyclase assays were performed on cells permeabilized by ‘swell dialysis’ as previously described (Voorheis and Martin, 1980; Rolin et al., 1998). Cells were incubated for 1 h at 4°C at a density of 5 × 108 cells ml−1 in TES buffer of low osmotic strength (50 mM KCl, 5 mM MgCl2, 1 mM glucose, 1 mM EGTA, 20 µg ml−1 leupeptine, 0.3 mM PMSF, 13.3 mM TES, pH 7.5). After swell dialysis, a sample of 107 cells (20 µl) was added to 80 µl of the assay cocktail (0.5 mM cAMP, 10 mM phosphocreatine, 50 units creatine kinase, 1 mM EGTA, 10 mM MgCl2, 20 mM KCl, 0.5 mM ATP, 20 µg ml−1 leupeptine, 25 mM TES pH 7.5 and 1 µCi of [α-32P]-ATP at 10–40 Ci mmol−1). The reaction was incubated for 5, 10 and 20 min at 37°C and was stopped by adding 100 µl of stop solution (2% SDS, 40 mM ATP, 0.01 M cAMP). Cyclic AMP was isolated according to Salomon et al. (1974) by two-step chromatography and measured by liquid scintillation counting. AC activity was calculated by linear regression analysis of the rate of cAMP production. Values were normalized to the total amount of protein present in each assay sample. Protein concentrations were determined by Bradford assay.
We acknowledge P. Poelevoorde for technical help and V. Ledent (ULB) for bioinformatic support. We thank G. Rudenko (Oxford) for a BAC clone, G. Cross (New York) for the pLew100v5 vector, P. Bastin (Paris) and K. Gull (Oxford) for the anti-PFR hybridoma, M. Brenndörfer (LMU) for tubulin qPCR primers and D. Tonn (LMU) for critical reading of the manuscript and language editing. This work was supported by the Interuniversity Attraction Poles Programme of Belgian Science Policy (E.P. and M.B.), DFG Grant 1100/7-1 (M.B.) and the University of Munich. D.S. was funded by a return grant from the Belgian Science Policy.