To assess the role of FGF15 in cortex development, we extended the analysis of Fgf15 expression in the developing telencephalon by in situ RNA hybridization, to complement previous reports 161719. At embryonic day 9.5 (E9.5), Fgf15 was expressed in the anterior forebrain neuroepithelium (Figure 1a), but was excluded from the anterior dorsal midline where Fgf8 was expressed (Figure 1a, b, arrowhead). At E12.5, Fgf15 was strongly expressed in the septum (Additional file 1g, arrow), preoptic area (Additional file 1h, arrow), and weakly expressed in the neuroepithelium of the PSB (Additional file 1g, arrowhead). In addition, Fgf15 was strongly expressed in the caudal ganglionic eminence (CGE; Additional file 1j, arrow).

The overlapping and complementary expression of Fgf15 and Fgf8 in the dorsal midline suggests regulatory interactions between these factors. Gain-of-function experiments by Gimeno and Martinez 16 demonstrated that FGF8 protein can induce Fgf15 expression in the mesencephalon and prosencephalon. Thus, we analyzed the expression of Fgf15 in severely hypomorphic Fgf8 mutants (Fgf8^Null/Neo) 22. At E9.5, Fgf15 expression was strongly reduced at the midbrain/hindbrain boundary (Figure 1c, d, arrows), consistent with Gimeno and Martinez 16. However, in the rostral forebrain, Fgf15 expression remained strong, and it was ectopically expressed in the anterior midline (Figure 1c, d, c', d', arrowheads). At E12.5, Fgf15 expression remained strong in the dysmorphic septal area and appeared to be increased in the CGE (Additional file 1e–j'). Thus, while Fgf15 is positively regulated by FGF8 in the midbrain, it is repressed by FGF8 in the prosencephalic midline.

There is evidence that SHH promotes Fgf15 expression in the diencephalon 35 and in the cortex 620. Our analysis of the Shh^-/- mutant extended these findings by demonstrating that Fgf15 expression in the rostral telencephalon requires SHH function (Additional file 2). Thus, while Fgf15 expression in the rostral telencephalon requires SHH, it remains strong in the severe Fgf8 hypomorph, despite the lack of telencephalic Shh expression in these mutants 22. In addition, Gli3, a repressor of the SHH signal, may to be over-expressed in the Fgf15 mutants at E12.5 (Additional file 2q, q'), suggesting a positive feedback between FGF15 and SHH.

To determine FGF15's functions in forebrain development we analyzed the Fgf15 null (Fgf15^-/-) mutant mouse 32, and focused on the cortical phenotype. Histological analysis showed that the mutant cortex was thinner than the wild type (Figures 2 and 3; quantification below). The gross morphology of Fgf15 mutant embryos is similar to the wild-type cohort (Additional file 3).

We started the analysis at E9.5, soon after neurulation has generated the telencephalic vesicles, by studying mRNA expression of genes known to have important roles in cortical development or known to be markers of FGF signaling.

CouTF1 promotes caudoventral cortical identity 363738. While reducing Fgf8 dosage increased CoupTF1 expression 2122, we observed that loss of Fgf15 strongly reduced its expression at E9.5 (Figure 1e, f). Subtle changes in expression are also observed, including increased Fgf8 expression (Figure 1g, h); however, establishing these changes would require more quantitative methods. On the contrary, we did not observe a major effect on the expression of two genes that are positively regulated by FGF signaling, Spry2 (Figure 1i, j) and Erm (Figure 1k, l). Expression of cMyc, a gene downstream of the Ras/MAPK signaling 3940, may be slightly increased (Figure 1m, n).

Next, we interrogated FGF15 function by analyzing CoupTF1, Mest, Sp8, Pax6, Emx2 and Erm expression in the cortical ventricular and mantle zones at E12.5 and E14.5. As at E9.5, CoupTF1 expression was strongly reduced (Figure 2a, b and 2m, n), in contrast with its increase in Fgf8 mutants 2122. Fgf15 and Fgf8 mutants also showed the opposite effects on Mest and Sp8 expression (Figure 2c, d and 2e, f, respectively) 2241.

On the other hand, Pax6, Erm and Emx2 showed subtle changes in the Fgf15 mutant. Pax6 cortical ventricular zone expression showed a modest decrease at E12.5 (Figure 2g, h) as well as Erm at E12.5 and E14.5 (Figure 2k, l and 2q, r). Emx2 expression may be increased in the dorsomedial cortical ventricular zone at E12.5 and E14.5 (Figure 2i, j and 2o, p). At this stage Fgf15 expression in the rostral patterning center does not appear to overlap with the cortical expression of the gene analyzed here. The changes in cortical expression could, however, result from the earlier patterning changes observed at E9.5 (Figure 1) and through diffusion of FGF15.

Zebrafish treated with morpholinos to Fgf19 (the name of zebrafish Fgf15) exhibit loss of subpallial molecular features 34. We, however, did not detect a strong subpallial phenotype in the mouse Fgf15 mutant, although there was a suggestion of reduced GAD67,Nkx2.1 and Pax6 expression in the mantle zone of the ventral septum and piriform cortex (Additional file 4). There was, however, a clear reduction in the number of GAD67⁺cortical interneurons in the dorsomedial cortex (Additional file 4a, b). Reelin expression in this region appeared normal showing that differentiation of the marginal zone was not grossly disrupted (Additional file 4e, f).

In summary, reduced expression of Fgf15 and Fgf8 has opposite effects on the expression of several transcription factors that have prominent functions in regulating cortical regional fate, proliferation and differentiation. This was particularly clear for CoupTF1, which promotes caudoventral cortical fate 3738, and which represses proliferation and promotes differentiation 38.

The cortex in the Fgf15 mutants was thinner than the wild type at E14.5 (Figures 2 and 3), suggesting that FGF15 may regulate the balance between proliferation and differentiation. To explore this, we analyzed markers of differentiation and indices of proliferation.

Neurogenin2 (Ngn2) expression marks progenitor cells in the cortical ventricular zone (VZ) and subventricular zone (SVZ) and is largely excluded from the postmitotic neurons of the cortical plate 42 (Figure 3a). Fgf15 mutants maintained strong Ngn2 expression in the VZ, while its expression in the SVZ seemed to be reduced; however, the Ngn2 negative area was much thinner (Figure 3a, b), suggesting reduced cortical plate thickness. To confirm this observation, we studied expression of β-III-Tubulin, a marker of differentiating neurons. Indeed, the width of the cortical postmitotic zone was greatly reduced in the Fgf15 mutant (50% in ventral regions and 45% in dorsal regions; Figure 3c, d), while the width of the VZ was larger in the mutants (30% in ventral regions and 25% in dorsal regions; Figure 3c, d). The reduced thickness of the cortical plate was confirmed by analyzing expression of markers of early born cortical projection neurons, Tbr1 and Mef2C (Figure 3e–h). Tbr1 and Mef2C expressions were reduced in the cortical plate of the Fgf15 mutants (Figure 3e, f, g, h). Thus, Fgf15 mutants show a defect in the growth of the cortical plate, perhaps due to defect in neurogenesis.

We explored the hypothesis that the cortical plate defect is caused because FGF15 may promote the switch between proliferation and differentiation. Thus, we assessed the mitotic index by immunostaining with phospho-histone H3 (PH3), a marker for M phase of the cell cycle. At E14.5, we observed a 37% increase in the density of PH3⁺ cells in the cortical VZ and a 42% increase in the SVZ (Figure 4a–c). This was already evident at E12.5 when the increase was 20% (Additional file 5a–c).

The rate of proliferation of the cortical precursor cells is linked to the length of the cell cycle; in addition, the fraction of cells in a given phase of the cell cycle is directly proportional to the length of that specific phase, relative to the total length of the cell cycle 4344. Sequentially exposing proliferating cells to iodo-deoxyuridine (IdU) and bromo-deoxyuridine (BrdU) allows one to differentiate between defined populations of proliferating cells and then to calculate the cell cycle length (see Materials and methods). We compared the length of the cell cycle in E12.5 and E14.5 wild type and Fgf15 mutants using IdU/BrdU double labeling 45 (Figure 4d–f; Additional file 5d–f). Cell cycle length in the mutants was reduced to 11 hours, compared to the 13.5 hours of the wild type at E14.5 (Figure 4d–f), and 10 hours, compared to the 11 hours of the wild type at E12.5 (Additional file 5d–f).

Next, we analyzed the fraction of neural precursor cells exiting the cell cycle to become neurons (Q fraction) at E14.5. To identify cells exiting the cell cycle, we used differential labeling by IdU and BrdU 45. The IdU/BrdU double labeling method allows one to distinguish between cells that are proliferating from cells that exit the cell cycle (see Materials and methods). Fgf15 mutants showed a 38% reduction in Q fraction (cells exiting the cell cycle) (Figure 4g–i).

Overall, these data provide evidence that reducing the dosage of Fgf15 reduces cell cycle length and cell cycle exit. Thus, FGF15 induces neurogenesis and this observation contrasts with the typical function of FGFs in promoting proliferation and maintenance of the neural progenitor state 14.

To investigate how FGF15 regulates cell cycle and neurogenesis, we studied signaling pathways that regulate the switch between proliferation and neurogenesis in Fgf15 mutants. First, we investigated the retinoic acid pathway, which promotes neural differentiation 46. We introduced an allele that expresses LacZ under the control of retinoic acid receptor enhancer elements 47 into the Fgf15^-/-background. At E12.5 and E14.5, there was a strong decrease in β-galactosidase activity in the mutants (Figure 5a–d), suggesting reduced retinoic acid signaling, consistent with the decrease in neuronal differentiation (Figure 3).

Wnt signaling through the β-catenin pathway promotes cortical proliferation 2484950 and represses differentiation of cortical neural progenitors 51. We introduced a Wnt/β-catenin LacZ reporter allele 52 into the Fgf15^-/- mutants and observed increased numbers of β-galactosidase⁺ cells in their cortex at E12.5 and E14.5 (Figure 5e–h), consistent with an increase in Wnt signaling, and the increase in proliferation (Figure 4).

Towards understanding the mechanism of FGF15 function, we examined the effect of Fgf15 mutation on selected components of the FGF signal transduction pathway in the rostral telencephalon at E9.5, E12.5 and E14.5. We focused on expression of Fgf receptors (Fgfrs) and Sprouty2 (Spry2); Sprouty genes encode negative regulators of the FGF signaling pathway that are induced by FGF8 245354.

At E9.5 and E12.5, expression of Fgf8 and Spry2 in the rostral patterning center and the rostral midline (that is, anlage of the septum) appeared normal in the Fgf15 mutants (Figures 1g–j and 6a–l). At later stages, E12.5 and E14.5, Spry2 expression was reduced in the VZ of the ventrolateral cortex (Figure 6e–h, arrows) while its expression in the dorsomedial cortex was preserved (Figure 6e–f, arrowheads and Figure 6i–l).

Next we examined expression of Fgfr1-4 at E12.5 and E14.5. Fgfr2 showed the largest change in expression: it increased in the VZ, particularly in the dorsolateral cortex (Additional file 6e–h); in the cortical plates its expression was lost. Fgfr1 showed subtle changes in the VZ (a slight increase), but was reduced in the cortical plate and intermediate zone (Additional file 6a–d). Fgfr3 expression was reduced in the dorsal cortex at E14.5 (Additional file 6i–l, arrows).

FGF15 preferentially binds to FGFR4 in vitro 33555657. We confirmed that Fgfr4 expression was not detectable in the E12.5 cortex (data not shown), as previously published 58. However, we did observe its expression in the superficial cortical plate at E14.5; this expression domain was preserved, but thinner in the Fgf15 mutants (Additional file 6m–n). Thus, Fgf15 mutants showed varied effects on the expression of several FGF receptors in cortical progenitors and in the cortical plate.

Toward establishing how FGF15 and FGF8 differentially affect cortical patterning proliferation and differentiation, we studied the effects of recombinant FGF15 and FGF8 proteins on dissociated E12.5 cortex; at this age most of the cortical cells are neuroepithelial progenitors.

Unlike mouse and human FGF2 and FGF8, which share 93% and 98% amino acid identity, mouse FGF15 shares only 51% identity with its human homolog FGF19 (data not shown) 2829. Thus, we compared the effect of purified mouse recombinant FGF15 protein with commercially available recombinant human FGF19 protein. Our results with FGF15 and FGF19 were indistinguishable (data not shown).

FGFs signal through two major pathways that when activated phosphorylate ERK (42/44) and the AKT 1459. These pathways can also activate other important kinases that regulate processes such as protein translation (S6 kinase) and Wnt signaling (GSK3 kinase).

First, we began titrating the FGFs to find concentrations that led to robust changes in the levels of pERK (42/44). We compared 5 ng/ml with 50 ng/ml and found that they led to similar results, with the higher concentration yielding easily detectable levels of pERK (42/44) (Figure 7a, b; Additional file 7c, d). We chose to focus on 50 ng/ml.

We found several qualitative differences in the response to FGF8 and FGF15. First, while both ligands induced a rapid (5 minutes) increase in pERK, their levels decreased more rapidly in the FGF15-treated samples (Figure 7a, b). We used recombinant FGF2 as a control, because it is a known mitogen for cortical progenitors 6061, and as an activator of the ERK kinase pathway (Additional file 7). FGF2 showed similar results as FGF8 (Figure 7; Additional file 7).

While both FGF8 and FGF2 induced the phosphorylation of AKT (pAKT; Figure 7b, and Additional file 7a, respectively), FGF15 did not (Figure 7a). Activation of ERK and AKT, through phosphorylation, results in the phosphorylation and subsequent inhibition of GSK3 activity 6263. FGF8 and FGF2 increased levels of pGSK3 (Figure 7d, and Additional file 7b, respectively), whereas FGF15 treatment did not increase pGSK3 (Figure 7c). As GSK3 is an inhibitor of the Wnt/β-catenin pathway 64, these results suggest that FGF2/8 may repress Wnt signaling more than FGF15 through this mechanism.

FGFs promote protein translation and cell growth through the mTor/S6 pathway 6566. FGF8- and FGF2-treated cells increased phosphorylation of S6 protein, albeit with slower kinetics than pERK and pAKT (Figure 7d, and Additional file 7b, respectively). On the contrary, FGF15 did not show this increase; in fact, we observed a reduction of S6 phosphorylation in the first 15 minutes (Figure 7c).

Finally, we compared the effects of FGF15 and FGF8 recombinant proteins on proliferation and differentiation of cortical progenitors in vitro. We treated primary E12.5 cortical cultures grown in vitro for 24 and 48 hours with either FGF15 or FGF8. While FGF8 induced an approximately 1.7-fold increase in the mitotic index (assessed by comparing the ratio of PH3⁺/total cells in FGF8-treated and untreated cortical cultures), FGF15 caused an approximately 0.25-fold reduction in the mitotic index (Figure 7e–j, and quantification in 7k). This result is consistent with the increased proliferation of the cortical progenitors we observed in the Fgf15 mutants, and may reflect the distinct effects of FGF15 on the phosphorylation of ERK (42/44), AKT, GSK3 and S6. We did not test these cultures for apoptosis; however, we have not observed a change in the number of apoptotic cells in the Fgf15 mutants (data not shown). A more detailed analysis is required to determine the role of FGF15 on neural progenitor apoptosis.

The vertebrate rostral patterning center expresses multiple Fgf genes: 3, 8, 15, 17 and 18 91011121314161967. It is unknown what induces expression of the earliest Fgf (Fgf8) in the anterior neural ridge. However, Fgf8 is positively upstream of Fgf17 and Fgf18 (Additional file 1q-r') 17, whereas it appears to repress Fgf15 in the rostral midline (Figure 1c, d). In the Fgf8^Null/Neosevere hypomorph, Fgf15 expression remains strong, albeit in a reduced area (Figure 1c, d), perhaps secondary to the reduced proliferation in this region 22. On the other hand, induction/maintenance of Fgf15 expression clearly depends on SHH function (Additional file 2) 62035, as does maintenance of Fgf8 expression 68. Fgf15 does not appear to have a major role in regulating expression of Spry2, an intracellular antagonist of FGF8 signaling (Figures 1g, j and 6a–l). On the other hand, there is maybe a subtle increase in Fgf8 expression in the Fgf15 mutant (Figure 1g, h); a more quantitative analysis will be needed to verify this observation. Thus, the rostral patterning center has parallel signaling capabilities, one dominated by FGF8 and the other dominated by FGF15, which depends on SHH and not FGF8 (see schema in Figure 8). Currently, functions for FGF3 and FGF18 in the mammalian rostral patterning center are not known, whereas FGF3 and FGF8 have redundant functions in zebrafish 67.

FGF8 and FGF17 promote rostral telencephalic fate, in part by repressing expression of CoupTF1, Fgfr3 and Emx2, and promoting expression of Er81, Erm, Pea3, Mest and Sp8 12172122242541. The dorsal patterning center, through its expression of BMPs and Wnts 1269, can, in principle, repress the rostral patterning center. BMPs can repress Fgf8 expression 68; BMP and Wnt positively regulate Emx2 expression 70 and reduced Emx2 leads to expansion of Fgf expression 121769. However, currently, in vivo genetic studies have principally linked Wnt signaling to hippocampal specification 12 and BMP signaling to choroids plexus specification 7172, so it remains unclear to what extent the dorsal patterning center participates in neocortical patterning.

Here we show that FGF15 functions, at least in part, to repress rostral telencephalic fate, through promoting CoupTF1 and repressing Mest and Sp8 (Figure 2). Thus, FGF15, in conjunction with the putative role of the dorsal patterning center, participates in repressing the size, and perhaps the nature, of the rostral telencephalon.

FGF8 is also essential for inducing the ventral telencephalon, through promoting expression of Nkx2.1 and Shh 2273. While Fgf15 is expressed more ventrally than Fgf8 within the rostral patterning center (septum) (this paper) 17, to date we have detected only mild subpallial or septal hypoplasia (Additional files 3 and 4; data not shown). On the other hand, reduced Fgf19 expression in zebrafish (homologue of mouse Fgf15), results in greatly reduce expression of subcortical molecular markers 34, perhaps suggesting divergent functions for these genes in fish and mammals. In the mouse Fgf15 mutant, there are reduced numbers of cortical GABAergic interneurons in the dorsomedial cortex (Additional file 4); at this point we can not distinguish whether this is a defect secondary to the abnormal cortical environment, or secondary to abnormal subpallial development (cortical interneurons are generated in the basal ganglia anlage).

We propose that Fgf15 expression in the rostral patterning center has a profound role in telencephalic patterning during early neurulation based on the reduction of CoupTF1 expression at E9.5 (Figure 1e, f). At this stage, the distribution of Fgf15 mRNA is in a rostrocaudal gradient whose pattern is roughly complementary to the expression of CoupTF1 mRNA (Figure 1a, c; Additional file 1a–d). Assuming that the RNA and protein distributions are similar at E9.5, then FGF15 secretion could promote CoupTF1 expression in the most rostral domain. At later stages, Fgf15 expression in the septum and CoupTF1 expression in the cortex are far apart; therefore, it seems unlikely that loss of Fgf15 from the septum at these stages contributes to CoupTF1 reduction, especially in the dorsolateral cortex. However, Fgf15 is also expressed at the PSB, and in the CGE (Additional file 1g–j); Fgf15 from those sites could contribute to maintaining CoupTF1 cortical expression at later developmental stages. This suggests that FGF15 could play a key role in the putative patterning center at the PSB. This is the first evidence for a role of a secreted factor expressed in this putative patterning center 78. Fgf15 is also expressed in the lateral and medial ganglionic eminences (LGE/MGE) and the thalamus; currently, we have not detected phenotypes in these structures.

The most overt phenotype in the Fgf15 mutants was reduced neurogenesis and increased proliferation in the cortex (Figures 3 and 4). Consistent with this, in vitro cultures of cortical progenitors with recombinant FGF15 have reduced neuronal proliferation (Figure 7). These data differ from the results obtained by reducing expression of the zebrafish Fgf15 homologue (Fgf19) with antisense oligonucleotides (morpholinos) 34. Whereas reduced Fgf19 expression led to a decrease in the number of proliferating cells in the embryonic fish brain, Fgf15^-/- mice showed increased proliferation. These divergent phenotypes could be explained by the different methodologies to reduce gene expression (constitutive deletion mutation versus morpholino), different compensatory responses, or distinct functions of mouse Fgf15 and zebrafish Fgf19. It is conceivable that the Fgf15 deletion produces an amino-terminal fragment with altered signaling properties; however, the deletion does remove exon 3, which encodes the residues predicted to bind heparin sulfate proteoglycans and FGF receptors 32. Future studies with a Fgf19 zebrafish mutant will help resolve whether zebrafish Fgf19 and mouse Fgf15 indeed have divergent functions that may contribute to the divergent morphogenesis of the fish and mammalian telencephalons.

We suggest that an important aspect of Fgf15's functions is to promote expression of the CoupTF1 transcription factor, because the cortical phenotypes observed through changing Fgf15's expression levels mirror those observed when altering CoupTF1 dosage. Increased CoupTF1 represses proliferation and promotes neurogenesis, whereas loss of CoupTF1 promotes proliferation and represses neurogenesis 38. As loss of Fgf15 expression results in decreased CoupTF1 expression as early as E9.5 (Figure 1), we suggest that the Fgf15 mutant phenotype is caused, in part, by reduced CoupTF1 levels. However, because the neurogenesis phenotype in Fgf15 mutants is more severe than that of the CoupTF1 mutants, other factors must contribute.

Unlike FGF15, FGF8 represses CoupTF1 2122. Furthermore, FGF8 promotes proliferation in the developing telencephalon 22 (Figure 7), as does FGF2 14607475. Thus, while the in vivo functions of certain FGFs (FGF2 and FGF8) promote proliferation in the developing telencephalon, FGF15 has the opposite role. Therefore, FGF15 and FGF8 provide opposing signals from the rostral patterning center that control the balance of proliferation and differentiation. One can imagine how differential modulation of these signals during development, disease, or evolution will have profound effects on cortical size, thickness and regional fate.

FGF15 and FGF8/17 are believed to signal through several receptor tyrosine kinases (FGFR1-4) that are negatively modulated by several mechanisms. Spry1-4 encode FGF-induced cytoplasmic repressor of FGF-signaling 5376. While Spry expression is clearly reduced in the Fgf8 mutant forebrain 1722, we did not detect a reduction in Spry expression in the rostral patterning center/septum of the Fgf15 mutants (Figures 1i, j and 6e–l). Fgfr3 appears to have atypical signaling outputs. In fact, Fgfr3 mutant mice have increased proliferation and reduced differentiation of chondrocytes 7778, and pancreatic cells 79. Fgf8 mutants have increased Fgfr3 expression 21, whereas the Fgf15 mutants show the opposite phenotype (Figure 6). Thus, we propose that FGF15 is a secreted modulator of several cellular processes that are promoted by FGF8.

Towards elucidating the differential effects of reducing FGF15 and FGF8, we compared the effects of recombinant FGF15/19 and FGF8 on the levels of phosphorylated forms of ERK (42/44), AKT, S6 and GSK3 in primary cultures made from E12.5 mouse cortex (Figure 7). We demonstrated that while FGF15 phosphorylation of ERK kinase (42/44) was transient (approximately 15 minutes), FGF8 phosphorylation of ERK was sustained over the time of the experiment (1 hour). The duration of FGF-signaling is associated with distinct cellular responses 80. For example, in fibroblasts, sustained activation of ERK correlates with S phase entry 81828384, while in PC12 cells, sustained, but not transient, activation of ERK induces differentiation into sympathetic-like neurons 85868788.

FGF15 and FGF8 also had distinct effects on the levels of pAKT and pS6. While FGF8 increased the levels of both pAKT and pS6 phosphorylation, FGF15 did not increase pAKT, and appeared to reduce pS6. The differences observed in the phosphorylation of the effector kinases of the FGF signaling pathway may contribute to the opposite phenotype of the FGF15 and FGF8 mutants observed in vivo and in vitro.

The simplest model to account for the signaling differences between FGF8 and FGF15 would be that they differentially activate/repress FGF, or other, receptors. While studies have established the effects of distinct ligands and receptor combinations on mitogenesis and in vitro binding 14555789, definitive elucidation of the in vivo biochemistry is lacking, particularly for FGF15. Although FGF15/19 preferentially binds FGFR4 33555657, this receptor does not appear to be expressed in the forebrain until approximately E14.5 (Additional file 6). In addition, FGF15/19 binds Klotho-β, which acts as a co-receptor. We have not observed any expression of Klotho-β in the forebrain at E12.5 and E14.5 (data not shown). Therefore, it is likely that FGF15 functions by regulating other FGF receptors. Therefore, additional analysis is needed to establish how FGF8 and FGF15 signal to elucidate how they differentially regulate the balance between proliferation and differentiation.

The Fgf15^-/- 32, Fgf8^Null/+ and Fgf8^Neo/+ 22, Shh^± 90, BAT-gal 52 and RARE-LacZ 47 strains were maintained on a mixed C57BL6/CD1 genetic background. Screening of the mutant alleles was performed by PCR genotyping as described previously 223290. Noon on the day of the vaginal plug was considered as E0.5. Mouse colonies were maintained at the University of California, San Francisco, in accordance with National Institutes of Health and UCSF guidelines.

Pregnant females were deeply anesthetized with CO₂ and sacrificed by cervical dislocation. The embryos were dissected and the brains were fixed by immersion in 4% paraformaldehyde in phosphate buffered saline. The tissue was cryoprotected by immersion in 30% sucrose/phosphate buffered saline, embedded in OCT (Tissue-Tek, Sakura Finetek, Torrance, CA, USA), and cryostat sectioned (10–20 μm).

Immunofluorescence on cryostat sections was performed as previously described 22. The antibodies used were as follows: mouse anti-Tuj1 (1:1,000; Covance, Princeton, NJ, USA), rabbit anti-PH3 (1:200; Upstate/Millipore, Billerica, MA, USA), mouse anti-BrdU (1:100; Becton Dickinson, Franklin Lakes, NJ, USA), rat anti-BrdU (1:100; Abcam, Cambridge, MA, USA). Goat anti-rabbit, goat anti-mouse and goat anti-rat secondary antibodies, conjugated with either Alexa 488 or Alexa 594 (1:300, Molecular Probes/Invitrogen, Carlsbad, CA, USA), were used at a dilution of 1:300.

In situ RNA hybridization on cryostat sections was performed as previously described 22. Comparison of gene expression changes between brains of different genotypes was performed by matching the planes of section to the best of our abilities, using multiple anatomical features. Whenever possible, this was performed for multiple planes of section for each gene, and from at least two brains for each genotype.

For the cell cycle kinetic analysis, a single injection of IdU (50 μg/g; Sigma, St Louis, MO, USA) was administered to pregnant females, carrying either E12.5 or E14.5 embryos, at T = 0. This was followed at T = 1.5 hours by a single injection of BrdU (50 μg/g; Sigma). Mice were sacrificed at T = 2 hours.

For the quantification of the numbers of cells exiting the cell cycle (Q fraction), IdU was administered as a single injection at T = 0. This was followed at T = 1.5 hours by sequential injections of BrdU every 3 hours, for a total of 15 hours. Mice were sacrificed at 0.5 hours after the last injection.

For the calculation of BrdU/IdU labeling index, we sampled two 100 μm bins spaced 200 μm apart in the ventro-lateral region of the cortex (rectangles in Figure 4 and Additional file 5). Images were acquired using a confocal microscope (Radiance 2000, Bio-Rad, Hercules, CA, USA) with a 20× or a 40× objective. BrdU/IdU positive cells (red channel) and BrdU positive cells (red channel) were counted together with the total number of cells for each bin (calculated by the number of Hoechst labeled nuclei). The experimenter was blind during sampling, image analysis, data collection and statistical analysis. Digitized images were imported into Phostoshop CS3 (Adobe) for counting. Statistical analysis was performed with SPSS (SPSS) and data plotted using Excel (Microsoft) and the significance level was taken as p < 0.05.

A total of three cases were counted in sections spaced evenly through the rostrocaudal cortex. The cell cycle length was calculated using the paradigm described in 91.

The adenovirus containing the Fgf15 coding region with a (His)₆ tag was the gift of Dr Steven Kliewer (UT Southwestern Medical Center, Dallas, TX, USA). The FGF15 adenovirus was grown in HEK293 cells (ATCC: CRL 1573) and titrated onto the same cells. 1.2 × 10⁷ COS7 cells (ATCC: CRL 1673) were plated in T150 flasks (6 flasks for each protein preparation). The cells were infected with adenovirus at a multiplicity of infection of 2 for 2 hours in Dulbecco's modified Eagle's medium (DMEM)/2% fetal calf serum. After the infection, the medium was changed with DMEM without fetal calf serum. Three days post-infection the supernatant was collected, equilibrated with a 10× solution of 500 mM Tris/HCl pH8, 1 M NaCl, 100 mM Imidazole, centrifuged 20 minutes at 10,000 g, and passed through a 0.8 ml nickel agarose column (Ni-NTA Agarose, Quiagen, Valencia, CA, USA). The column was washed with 25 column volumes of 50 mM Tris/HCl pH 8, 0.25 M NaCl, 0.1% NP-40, 10 mM Imidazole, and then with 25 column volumes of 20 mM Tris/HCl pH 8, 0.25 M NaCl, 10 mM Imidazole. The His-tagged FGF15 protein was eluted in 20 mM Tris/HCl pH 8, 0.25 M NaCl, 0.5 M Imidazole. The eluate was passed through a PD MidiTrap G-25 desalting column (GE Healthcare, Piscataway, NJ, USA) following the manufacturer's protocol, eluted in 10 mM Tris/HCl pH 7.5, 135 mM NaCl, 5 mM KCl, 1 mM MgCl₂, and finally concentrated with the Amicon Ultra-4 3 k centrifugal filter device (Millipore, Billerica, MA, USA). The purity of the protein preparation was assessed by Coomassie Blue staining and by immunoblotting (western blot) using an anti-FGF15 polyclonal antibody (SC 16816, Santa Cruz Biotechnology, Santa Cruz, CA, USA).

Wild-type CD1 brains (E12.5) were used to prepare primary cortical cultures. The brains were removed in ice-cold Hank's solution (HBSS; Invitrogen, Carlsbad, CA, USA), and the cortices dissected, after removing the external membranes. The samples were mechanically dissociated in DMEM/10% fetal bovine serum (FBS; Invitrogen) and the single cell suspensions were plated at a density of 2.5 × 10⁵ cells per cm² in two-well chamber slide (Lab-Tek, Nalgene, Rochester, NY, USA). We were careful to plate the cells at the same density in all of the wells. After 16 hours the medium was replaced with DMEM/5% FBS and finally changed with DMEM/0.5% FBS after 24 hours. The cells were starved in DMEM/0.5% FBS for 24 hours before the treatment with the recombinant proteins. The mouse FGF15 recombinant protein was added to the cells at a concentration of 50 ng/ml in DMEM/0.5% FBS/5 μg/ml Heparin. This protein concentration corresponds to 1.7 pmol/ml for FGF2 and 2 pmol/ml for FGF15 and FGF8. The same protocol was used for human FGF2, FGF8 and FGF19 recombinant proteins (Fitzgerald Industries International, RDI, Concord, MA, USA).

For counting the cultured cells, we used an unbiased method that gave every cell an equal chance of being sampled. We defined a systematic series of fields of view in the culture area to cover the surface of the well. The experimenter was blind to sample identity during sampling, image analysis, data collection and statistical analysis. Images were acquired on a fluorescence microscope with a 20× objective. We sampled at least two different wells for each time point obtained from four independent experiments. Digitized images were imported into Phostoshop CS3 (Adobe) for counting. Statistical analysis was performed with SPSS (SPSS) and data plotted using Excel (Microsoft) and the significance level was taken as p < 0.05.

Cells from primary E12.5 cortical cultures were lysed in radioimmunoprecipitation buffer (RIPA: 50 mM Tris/HCl pH 7.5, 150 mM NaCl, 0.5% sodium deoxycolate, 0.1% SDS, 1% NP-40) supplemented with protease inhibitors (Protease Inhibitor Cocktail, Roche, Indianapolis, IN, USA) and phosphatase inhibitors (Phosphatase Inhibitor Mix 1 and 2, Sigma). Protein samples (10 μg each) were resolved on polyacrylamide gels and transferred to nitrocellulose membranes by electroblotting. Membranes were pre-incubated in 5% nonfat dry milk in Tris buffered solution (TBS)/0.1% Tween-20 (TBST) and then incubated O/N with the primary antibodies in TBST/5% bovine serum albumin in TBST. The primary antibodies were: rabbit-anti phospho-p44/42 (Thr202/Tyr204) MAPK (1:1,000; Cell Signaling, Boston, MA, USA), rabbit anti-phospho-Akt (Ser 473) (1:500; Cell Signaling), rabbit anti-phospho-GSK3α/β (Ser 21/9) (1:1,000; Cell Signaling), rabbit anti-phospho-S6 ribosomal protein (Ser235/236) (1:1,000; Cell Signaling), and mouse anti-α-tubulin (1:10,000; Sigma). Membranes were probed with the appropriate goat horseradish peroxidase-conjugated antibodies (1:500; Biorad, Hercules, CA, USA) and developed using ECL reagents (GE Healthcare). Each experiment was repeated at least three times. Quantification of optical density was done using NIH ImageJ software; the intensity of the bands was normalized with respect to the intensity of the α-Tubulin band. The numbers indicated in Figure 7 and Additional file 7 represent the fold-induction, or reduction, from the time point before FGF addition (T0).