Corticotrophin-releasing factor type 2 receptor-mediated suppression of gonadotrophin-releasing hormone mRNA expression in GT1-7 cells

Abstract

Corticotrophin-releasing factor (CRF) released during stress has been implicated in the suppression of the hypothalamo–pituitary–gonadal (HPG) axis, especially the gonadotrophin-releasing hormone (GnRH) pulse generator, the central neural regulator of pituitary LH and FSH secretion, resulting in reproductive dysfunction. The gonadal steroid 17β-oestradiol (E₂) has been shown to enhance CRF- and stress-induced suppression of pulsatile LH secretion. In the present study, we investigated the potential direct action of CRF on GnRH neurones by using GT1-7 cells, an established GnRH cell line. Furthermore, we investigated the modulatory influence of E₂ on the effects of CRF and expression of CRF type 2 receptors (CRF-R2). Expression of CRF-R2 in the GT1-7 cells was detected by reverse transcription-polymerase chain reaction (RT-PCR). CRF produced a dose-dependent suppression of GnRH mRNA expression, an effect reversed by the selective CRF-R2 antagonist, astressin₂-B (Ast₂-B). E₂ combined with CRF resulted in a greater suppression of GnRH expression compared with either treatment alone. E₂ also increased CRF-R2 expression. These results demonstrate for the first time expression of CRF-R2 in the GT1-7 cells and suggest that CRF may directly regulate GnRH gene expression, an effect mediated, at least in part, by CRF-R2. They also raise the possibility that up-regulation of CRF-R2 may contribute to the sensitising influence of E₂ on CRF- and stress-induced suppression of the GnRH pulse generator.

• CRF
• CRF-R2
• stress
• GT1-7
• GnRH

Introduction

Corticotropin-releasing factor (CRF) is the principal hypophysiotrophic factor driving the hypothalamic–pituitary–adrenal (HPA) axis. CRF is a member of a related family of peptide ligands including urocortin (Ucn) I, UcnII and UcnIII. The physiological actions of these peptides are mediated by interactions with the two major CRF receptor subtypes, CRF type 1 receptors (CRF-R1) and CRF type 2 receptors (CRF-R2) (Perrin et al. 1999). Both CRF receptor subtypes are broadly distributed throughout the central nervous system, with heterogeneous anatomical distribution patterns observed (Chalmers et al. 1995). CRF-R1 expression is most abundant in the brainstem, cerebellum, amygdala and cortex, whereas CRF-R2 expression is more prevalent in sub-cortical structures, including the lateral septum and various hypothalamic nuclei (Chalmers et al. 1995). CRF-R1 has been shown to play a crucial role in stress-associated responses of the HPA axis and anxiety-related behaviour (Timpl et al. 1998). It has been reported that CRF-R2 may have a counteracting role in the HPA axis stress response (Bale et al. 2000; Coste et al. 2000). However, there is evidence for both anxiolytic (Bale et al. 2000; Kishimoto et al. 2000) and anxiogenic (Takahashi et al. 2001) roles for CRF-R2.

It is well established that CRF is a potent suppressor of the gonadotrophin-releasing hormone (GnRH) pulse generator, the central regulator of pituitary LH and FSH secretion (Knobil 1992). Central administration of CRF results in inhibition of LH pulses in rats (Rivier and Vale 1984; Cates et al. 2004) and monkeys (Gindoff and Ferin 1987; Williams et al. 1990). Moreover, central infusion of non-selective CRF receptor antagonists can reverse the LH pulse-suppressing effects of a variety of stressors including footshock (Rivier et al. 1986), fasting (Maeda et al. 1994), glucoprivation with 2-deoxyglucose (Tsukahara et al. 1999) or hypoglycaemia (Chen et al. 1996; Cates et al. 2004). However, study of the role played by the CRF receptor subtypes in mediating stress-induced suppression of the GnRH pulse generator has only recently been initiated. We have shown that the CRF-R2 antagonist, astressin₂-B (Ast₂-B), completely blocked restraint stress-induced suppression of pulsatile LH secretion and that central administration of the CRF-R2 specific ligand Ucn II induced a dose-dependent suppression in LH-pulse frequency in the rat (Li et al. 2005). Further, we have shown that Ast₂-B markedly attenuated the LH pulse-suppressing effects of insulin-induced hypoglycaemia and innate immunological challenge with lipopolysaccharide (LPS), while the latter two responses were not blocked by selective CRF-R1 antagonists (Li et al. 2006). These data suggest a pivotal role of CRF-R2 in the control of the GnRH pulse generator.

The precise mechanism by which CRF affects the GnRH neural system remains to be fully elucidated. Synaptic contacts have been observed between CRF axon terminals and GnRH neurones in the medial preoptic area of rats (MacLusky et al. 1988). More recently, contacts have been observed between CRF fibres and GnRH perikarya in the medial preoptic area and infundibulum of humans (Dudas and Merchenthaler 2002; Dudas and Merchenthaler 2006), thus providing an anatomical substrate for a direct functional action of CRF on the GnRH neuronal system. Although the expression of CRF-R1 has been shown, both in vivo, by microarray analysis (Todman et al. 2005) and single-cell reverse transcription-polymerase chain reaction (RT-PCR) from GnRH neurones (Jasoni et al. 2005) and in vitro, by CRF-R1 immunoreactivity in the Gn11 cell, a GnRH cell line (Tellam et al. 1998), the presence of CRF-R2 in the GnRH neurone is yet to be identified.

The gonadal steroid, 17β-oestradiol (E₂) has been shown to potentiate the suppression of pulsatile LH secretion in response to a variety of stressors in several species (Cagampang et al. 1991; Chen et al. 1992; Cagampang et al. 1997; Tsukahara et al. 1999; Li et al. 2003). Furthermore, E₂ has been shown to enhance the inhibitory effects of CRF on pulsatile LH release (Cates et al. 2004). The mechanisms by which E₂ exerts its sensitising effect on both stress- and CRF-induced suppression of LH pulses remains to be fully elucidated. However, E₂ may mediate its sensitising effects directly on the GnRH neurones themselves, since the presence of oestrogen receptor (ER)-β immunoreactivity (Hrabovszky et al. 2001; Kallo et al. 2001) and the presence of ER-β but not ER-α mRNA (Hrabovszky et al. 2000) has been detected in GnRH neurones of rats. However, expression of both ER-α and ER-β has been detected in the GnRH neuronal GT1-7 cell line (Roy et al. 1999).

The aims of the present study were to identify expression of CRF-R2 in the established GnRH GT1-7 cell line (Mellon et al. 1990), to test the hypothesis that CRF can act directly on GnRH cells to suppress GnRH mRNA expression and to investigate the role of CRF-R2 in this response. Furthermore, we have investigated the modulatory influence of E₂ on the effects of CRF and expression of CRF-R2 in the GT1-7 cell.

Materials and methods

Cell culture and reagents

The GT1-7 cells were cultured in DMEM supplemented with 10% FCS, 4.5 mg/ml glucose and penicillin/streptomycin and maintained at 37°C in an atmosphere of 5% CO₂ as described previously (Bowe et al. 2003; Kinsey-Jones et al. 2005). Cells were sub-cultured into multi-well plates 1–2 days before experiments and grown to approximately 80% confluence. The growth medium was removed and replaced with phenol red-free DMEM supplemented with 2% charcoal-stripped FBS and penicillin/streptomycin, 12 h prior to cell treatments. Cell treatments were added to this medium to give the intended concentration and left for the appropriate time before cell harvesting. CRF, α-helical CRF_9–41, Ast₂-B (Sigma-Aldrich Ltd., Poole, UK) were dissolved in sterile H₂O and made up to the final concentrations in medium. E₂ (Sigma-Aldrich) was dissolved in alcohol and then diluted to the final concentration in medium.

Total Rna Extraction, Rt-pcr, And Quantitative Rt-pcr Analysis

GT1-7 cells were harvested on ice using TRI reagent (Sigma-Aldrich) and then homogenized. The homogenates were stored at room temperature for 5 min to allow total dissociation of nucleoprotein and total RNA was then extracted following the manufacturer's protocol. Reverse transcription was then carried out using the reverse transcriptase M-MLV RT (Promega, Madison, WI, USA) following the manufacturer's protocol and using a combination of oligo dT primers and random primers.

For the PCR to identify CRF-R2, the following primers were used: (sense) 5′-ATGTTTGTGGAGGGCTGCTA-3′ and (antisense) 5′-GTCCACCAAATCACCAGCTT-3′, which span an intron and do not differentiate between CRF-R2α and CRF-R2β. The CRF-R2 sense primer corresponds to nucleotides 792–811 of the Genbank sequence (NM009953.1). PCR reaction conditions were 5 min at 95°C for one cycle, then 30 s at 95°C, 30 s at 60°C and 30 s at 72°C for 35 cycles. Each reaction included 2 μl of sample cDNA, 1 μl of 25 μM antisense and sense primers, 5 μl of 2.5 μM dNTPs, 5 μl Taq DNA polymerase buffer containing 15 mM MgCl₂, 0.5 μl Taq DNA polymerase (Promega). PCR product was sequenced and analysed using an ABI PRISM 310 (Applied Biosystems, Foster City, CA, USA).

For the quantitative PCR, the following primers were used: GnRH (sense) 5′-CTACTGCTGACTGTGTGTTTG-3′ (antisense) 5′-CATCTTCTTCTGCCTGGCT-3′; 28S (sense) 5′-TTGAAAATCCGGGGGAGAG-3′ (antisense) 5′-ACATTGTTCCA-ACATGCCAG-3′, CRF-R2 primers as before. The GnRH sense primer corresponds to nucleotides 1617–1637 of the GenBank sequence (M14872); the antisense ends at position 4535 of the GenBank sequence. These primer pairs flank intron–exon boundaries. The LightCycler^TM (Roche Biochemicals, Lewes, UK) was used for real-time quantitative analysis of GnRH mRNA expression. The sample cDNA prepared as above was used as a template for the PCR. During PCR, the amplified GnRH product was detected after each annealing phase in real time using the Faststart DNA Master SYBR Green I kit (Roche Biochemicals). Each reaction included 2 μl of sample cDNA (optimised so that sample values of the PCR product were within the standard curve), 0.5 μl of 25 μM antisense and sense primers, 2 μl 15 mM MgCl₂, 1 μl Faststart DNA master SYBR Green mix and 4 μl of water to give a total reaction volume of 10 μl. The GnRH reaction conditions were 10 min at 95°C for one cycle, then 10 s at 95°C, 10 s at 65°C and 10 s at 72°C for 35 cycles. The CRF-R2 reaction conditions were 10 min at 95°C for one cycle, then 10 s at 95°C, 10 s at 60°C and 10 s at 72°C for 35 cycles. The 28S reaction conditions were 10 min at 95°C for one cycle, then 15 s at 95°C, 10 s at 54°C and 5 s at 72°C for 28 cycles. Reaction conditions for GnRH, CRF-R2 and 28S were optimized separately to give the best results for each primer and for the different quantities of target in samples. Preliminary experiments were done to optimize the Mg^{2 +} concentration, to confirm PCR specificity by agarose gel electrophoresis and melting curve analysis and to prepare the PCR products used to generate standard curves in real-time PCR. GnRH and CRF-R2 were quantified, against a standard curve of samples containing known GnRH/28S PCR product concentrations, using the LightCycler^TM program. 28S rRNA was quantified as a reference gene against a separate standard curve of samples containing known concentrations of 28S product. The melting curves for GnRH, CRF-R2 and 28S generated by the LightCycler^TM program demonstrated that single products were amplified.

Statistical analysis

The effects of treatment were assessed on raw data using one-way ANOVA followed by Dunnett's test to determine significance before being converted to percentage values. Experimental sample numbers represent results from different RNA extracts obtained from different wells of GT1-7 cells treated identically. Results are given as the mean ± SEM percentage difference from the control.

Results

Expression of CRF-R2 in the GT1-7 cell

Expression of CRF-R2 mRNA was detected in the GT1-7 cell samples. Figure 1 shows expression of a single band at 192 bp; this band is identical to that observed in the mouse whole brain controls. DNA sequencing of these PCR products confirmed the identity as that of CRF-R2.

Figure 1 RT-PCR analysis of the CRF-R2 transcript present in GT1-7 cells. Expression of the expected 216-bp CRF-R2 product is present in the GT1-7 sample. Mouse brain tissue served as a positive control and water as a negative control. The experiment was repeated three times on independent samples. The amplified products were resolved in a 1% agarose gel and DNA fragments from a HpaII digest of pBluescriptII served as a molecular weight marker.

Effects of CRF on GnRH expression

GT1-7 cells were exposed to 10^{− 6} M CRF treatments for 6, 12, 24 and 48 h initially to determine the optimum exposure time (Figure 2(A)). A significant suppression (61.1 ± 6.8%) of GnRH mRNA expression was observed at the 6 h time point (P < 0.05, one-way ANOVA; Figure 2(A)). Although a tendency to suppress GnRH mRNA expression was observed at the other time points, significance was not reached. Consequently, all subsequent experiments were carried out at the 6 h time point. GT1-7 cells were treated with different concentrations of CRF ranging from 10^{− 6} to 10^{− 9} M (Figure 2(B)). CRF treatments induced a dose-dependent decrease in GnRH mRNA expression, with significant suppressions of 63.7 ± 6.9 and 51.5 ± 10.7% observed at 10^{− 6} and 10^{− 7} M, respectively (P < 0.05, one-way ANOVA; Figure 2(B)). The suppressive effects of CRF (10^{− 7} M) on GnRH mRNA expression were completely reversed by co-treatment with the non-selective CRF receptor antagonist α-helical CRF_9–41 (10^{− 5} M), which alone had no effect on GnRH expression (P>0.05, one-way ANOVA; Figure 3). To investigate the role of CRF-R2 in mediating the inhibitory effects of CRF on GnRH gene expression, the CRF-R2 selective antagonist Ast₂-B was used. The suppressive effects of CRF (10^{− 7} M) on GnRH expression were completely reversed by co-treatment with Ast₂-B (10^{− 5} M), which alone had no effect on GnRH mRNA expression (P>0.05, one-way ANOVA; Figure 3).

Figure 2 Effects of CRF on GnRH mRNA expression in GT1-7 cells at different time points and with different doses. (A) A significant suppression in GnRH mRNA expression was observed with 6 h CRF (10^{− 7} M) treatment. (B) The effect of 6 h treatment with different concentrations of CRF on GnRH mRNA expression in the GT1-7 cell line. CRF induced a dose-dependant inhibition of GnRH mRNA expression with the maximum response at 10^{− 6} M. Quantification of GnRH mRNA and 28S rRNA was carried out on all samples and the values expressed as a ratio of GnRH mRNA to 28S rRNA (mean ± SEM). The results are presented as the percentage of control. *P < 0.05 vs. control: One-way ANOVA, followed by Dunnett's test. All treatments were performed in triplicate and experiments were repeated three times.

Figure 3 Effect of CRF alone or in combination with the CRF receptor antagonists: α-helical CRF_9–41 (α-h CRF) or Ast₂-B, on GnRH mRNA expression in the GT1-7 cell line. The inhibitory effect of CRF (10^{− 7} M) on GnRH mRNA expression was reversed by both α-h CRF and Ast₂-B. Quantification of mRNAs for GnRH and 28S rRNA was carried out on all samples and the values expressed as a ratio of GnRH mRNA to 28S rRNA (mean ± SEM). The results are presented as the percentage of control. *P < 0.05 vs. control, CRF+α-hCRF, CRF + Ast₂-B, α-hCRF alone or Ast₂-B alone: One-way ANOVA, followed by Dunnett's test. All treatments were performed in triplicate and experiments were repeated 3–4 times.

Effect of CRF and E₂ co-administration on GnRH expression

Co-treatment with CRF and E₂ for 6 h resulted in a significant decrease (73.6 ± 6.8%) in GnRH mRNA expression in the GT1-7 cells (P < 0.05, one-way ANOVA; Figure 4(A)). This decrease in GnRH expression was significantly greater than that observed with 6 h CRF or E₂ treatments alone (Figure 4(A)).

Figure 4 (A) Effect of CRF and E₂ co-administration on GnRH mRNA expression in GT1-7 cells. The suppression of GnRH mRNA expression observed with co-treatment of E₂ and CRF was significantly greater than with CRF or E₂ alone. Quantification of GnRH mRNA and 28S rRNA was carried out on all samples and the values expressed as a ratio of GnRH mRNA to 28S rRNA (mean ± SEM). The results are presented as the percentage of control. *P < 0.05 vs. control, †P < 0.05 vs. CRF 10^{− 7} M, #P < 0.05 vs. CRF 10^{− 7} M + E₂ 10^{− 8} M: One-way ANOVA, followed by Dunnett's test. (B) The effects of E₂ on CRF-R2 mRNA expression in GT1-7 cells at different time points. A significant increase in CRF-R2 expression was observed with 6 h E₂ (10^{− 8} M) treatment. Quantification of CRF-R2 mRNA and 28S rRNA was carried out on all samples and the values expressed as a ratio of CRF-R2 mRNA to 28S rRNA (mean ± SEM). The results are presented as the percentage of control. *P < 0.05 vs. control: One-way ANOVA, followed by Dunnett's test. All treatments were performed in triplicate and experiments were repeated 3–4 times.

Effects of E₂ on CRF-R2 expression in the GT1-7 cell

To investigate the possible modulation of CRF-R2 expression by E₂, GT1-7 cells were exposed to 10^{− 8} M E₂ treatments for 6, 12 and 24 h. A significant increase (186.3 ± 36.2%) in CRF-R2 mRNA expression was observed at the 6 h time point (P < 0.05, one-way ANOVA; Figure 4(B)). Although a tendency to increase CRF-R2 mRNA expression was observed at the other time points, significance was not achieved.

Discussion

In the present study, we have shown for the first time that CRF induced a dose-dependent decrease in GnRH mRNA expression in GT1-7 cells, an effect which was reversed by the non-selective CRF receptor antagonist, α-helical CRF_9–41 and the CRF-R2 selective antagonist, Ast₂-B, indicating that this effect was CRF receptor-mediated and more specifically, may involve activation of the type 2 receptor. These data support previous in vivo studies showing that infusion of CRF induced a potent suppressive effect on the hypothalamic–pituitary–gonadal (HPG) axis, specifically on the GnRH pulse generator, thus decreasing pulsatile LH secretion in rats (Rivier and Vale 1984; Cates et al. 2004), monkeys (Gindoff and Ferin 1987; Williams et al. 1990) and humans (Barbarino et al. 1989). Indeed, CRF antagonists can block the suppression of the GnRH pulse generator by a variety of stressful stimuli. However, little is known about the mechanisms that underlie the pivotal role of CRF in stress-induced suppression of the GnRH neural system. The results from the present study raise the possibility that CRF may act at the level of the GnRH neurone. This is supported by evidence showing that CRF infusion into the GnRH neurone-rich medial preoptic area induced a decrease in both plasma LH levels (Rivest et al. 1993) and LH pulse frequency (personal observation). Further, an in vitro study using Gn11 cells, transfected with a GnRH promoter/luciferase reporter construct, demonstrated a decrease in GnRH promoter activity with CRF treatment (Tellam et al. 1998). However, future studies describing the release of GnRH from GT1-7 cells in response to CRF are required.

Expression of CRF-R1 has been shown in GnRH neurones, using green fluorescent protein promoter transgenic mice and oligonucleotide microarrays (Todman et al. 2005) and immunocytochemistry (Jasoni et al. 2005). We have detected CRF-R1 mRNA expression in GT1-7 cells (personal observation) and CRF-R1 immunoreactivity has also been detected in the Gn11 cell (Tellam et al. 1998). In the present study we have identified, for the first time, expression of CRF-R2 mRNA in the GT1-7 cell. Furthermore, our results indicate that this receptor is functionally active, since co-administration of Ast₂-B and CRF completely reversed the suppressive effects of CRF on GnRH mRNA expression. This suggests that the inhibitory action of CRF on the GT1-7 cells may be mediated, at least in part, by activation of CRF-R2. However, further work is required to determine protein levels for the receptor in the GT1-7 cells. We have recently shown that CRF-R2 plays a pivotal role in restraint (Li et al. 2005), LPS or hypoglycaemic (Li et al. 2006) stress-induced suppression of LH pulses in the rat. Furthermore, central administration of UcnII, a highly selective ligand for CRF-R2, induced a dose-dependent suppression of pulsatile LH secretion (Li et al. 2005). In light of the results from the present study, the crucial role played by CRF-R2 in mediating suppression of the GnRH pulse generator by a variety of stressful stimuli may possibly involve an action at the level of the GnRH neurone. The presence of synaptic connection between CRF axon terminals and GnRH neurones in rats (MacLusky et al. 1988) and humans (Dudas and Merchenthaler 2002, 2006), provides the anatomical substrate for a direct interaction between the CRF and the GnRH neural systems in situ. However, further work is required to examine whether CRF-R2 is expressed in the GnRH neurones in vivo, since definitive evidence for absence of the receptor is lacking (Todman et al. 2005). Furthermore, although GT1-7 cells have a neural phenotype, synthesise GnRH, release GnRH in a pulsatile fashion, and are widely used as an in vitro model for the GnRH neuronal system, it is important to appreciate the limitations of extrapolating findings in the present study to the physiological mechanisms underlying the operation of the GnRH pulse generator in vivo.

While we have only investigated the role of CRF-R2 in the present study, the potential role of CRF-R1 cannot be ignored and requires further study. Although expression of CRF-R1 mRNA was recently detected in mouse GnRH neurones (Jasoni et al. 2005; Todman et al. 2005), double-labeled in situ hybridisation of mRNAs for GnRH and CRF-R1 failed to localize CRF-R1 mRNA in GnRH neurones in the rat (Hahn et al. 2003). Nevertheless, we have recently shown that administration of a selective CRF-R1 antagonist reversed the inhibitory effects of restraint on LH pulses in the rat (Li et al. 2006).

Stress-induced suppression of pulsatile LH secretion is potentiated by the presence of E₂. This effect is observed in response to hypoglycaemia in rats (Cagampang et al. 1997; Li et al. 2003), monkeys (Chen et al. 1992) and sheep (Adam and Findlay 1997) and in response to fasting (Cagampang et al. 1991) and glucoprivation with 2-deoxyglucose (Tsukahara et al. 1999; Nagatani et al. 2001) in rats. Furthermore, we have recently shown that E₂ augments CRF-induced suppression of LH pulses in the rat (Cates et al. 2004). The results from the present study show that E₂ combined with CRF has a greater suppressive effect on GnRH mRNA expression in the GT1-7 cell compared with CRF or E₂ treatments alone. These results suggest that one possible site of action of E₂ to potentiate CRF- or, indeed, stress-induced disruption of the GnRH neural network, could be at the level of the GnRH neurone per se. GnRH neurones, including GT1-7 cells contain the necessary ERs (Roy et al. 1999; Hrabovszky et al. 2000; Hrabovszky et al. 2001; Kallo et al. 2001). In addition, there is considerable evidence that gonadal steroids increase expression of CRF mRNA in the hypothalamic paraventricular nucleus (PVN) (Bohler et al. 1990; Van Vugt et al. 1997; Li et al. 2003), potentially resulting in an increased readily releasable pool of CRF. However, the role of the PVN in mediating stress-induced suppression of the HPG axis is unclear, since complete electrolytic lesion of this nucleus does not abolish the suppression of LH release in response to either foot shock or interleukin-1 stress (Rivest and Rivier 1991). In the present study, we have shown that E₂ is able to increase expression of CRF-R2 mRNA in the GT1-7 cell and may therefore increase the sensitivity of the cell to CRF signalling, thus providing a potential mechanism through which E₂ sensitisation occurs. It is of note that the maximum increase in CRF-R2 mRNA expression at 6 h in response to E₂ coincided with the time point for maximum suppression of GnRH mRNA examined in response to CRF. It has previously been shown that the stress-induced increase in hypothalamic CRF receptor expression is further enhanced in proestrus rats, suggesting a facilitatory action of E₂ (Nappi and Rivest 1995). Furthermore, an oestrogen response element is present in the promoter region of CRF-R2 (Catalano et al. 2003), thus providing a potential mechanism of E₂ modulation of CRF-R2 expression in the GT1-7 cell.

In summary, the results of the present study indicate that CRF acts directly on the GT1-7 cells to suppress GnRH gene expression and therefore raise the possibility of a direct action of CRF on the GnRH neural system in vivo. Furthermore, these effects appear to be mediated by CRF-R2, and may therefore, underlie the pivotal role played by CRF-R2 in stress-induced suppression of the GnRH pulse generator. The action of E₂ to increase CRF-R2 expression may also contribute to the sensitising influence of this gonadal steroid on CRF- and stress-induced suppression of the GnRH neural system.

Acknowledgements

The authors thank Dr A. F. Parlow (NIDDK) for providing the LH RIA kit. The authors are grateful to Professor P.L. Mellon for providing the GT1-7 cells. This work was supported by the Wellcome Trust. J.E.B. is a recipient of the Guy's and St Thomas' Charitable Foundation (UK) PhD Studentship.