Neuroanatomy of the FGF system


  • Martin H. Dominguez,

    1. Department of Neurobiology and Kavli Institute for Neuroscience, Yale University School of Medicine, New Haven, Connecticut 06510
    Search for more papers by this author
  • Pasko Rakic

    Corresponding author
    1. Department of Neurobiology and Kavli Institute for Neuroscience, Yale University School of Medicine, New Haven, Connecticut 06510
    • Yale University School of Medicine, P.O. Box 208001, 333 Cedar Street, New Haven, CT 06520-8001
    Search for more papers by this author

The papers in the Journal of Comparative Neurology used to contain complicated neuroanatomiacal diagrams with acronyms that annoyed molecular biologists, such as, “PF (prefrontal cortex) is reciprocally connected with MD (medial dorsal nucleus) and projects to the NS (neostriatum) and PGN (pontine gray nuclei) of the RB (rombencephalon),” with the arrows and T-shaped lines indicating, respectively, excitatory and inhibitory influences. Now, molecular biologists have their revenge with such diagrams where FGF8 is connected reciprocally with Sp8, BMPs, WNTs, Emx2, and one way to Foxg1 (Fig. 1). Still, we all have to learn to live with an ever-growing number of acronyms, because it would not be possible, in the limited space available, to explain complex neuronal or molecular interactions without them. The instructive case is the development of the mammalian neocortex that is both bolstered and cursed by its seemingly innumerable dimensions that make it exciting, but also challenging, to understand and investigate.

Figure 1.

Schematic showing the molecular interactions between FGF and up- and downstream molecular pathways based on the numerous studies and reviews that were cited in this commentary. Of particular interest is the mutual repression between caudal and rostral patterning centers; this inhibition directs and shapes the ultimate cortical landscape, and has been leveraged by Cholfin and Rubenstein as well as many similar studies.

The most obvious dimensions of analysis include timing, sequence, genesis of phenotypes, columns, layer formation, and evolution of species-specific differences in areal parcellation. The latter of these, the functional area assignments across the cortical landscape, are often ignored in otherwise complex studies for the sake of simplicity. Although the radial unit and inside-out paradigms hold true throughout developing dorsal telencephalon (Rakic,1988), the specifics are likely fine-tuned and adjusted in an area-specific fashion from as early as the birth of the preplate (Fukuchi-Shimogori and Grove,2003) or even before the onset of neurogenesis (Bystron at al.,2008), the essence of the protomap theory (Rakic,1988). Ongoing work continues to show pronounced differences in the developmental scheme of neocortex when several unique areas are examined. The manifestations of such areal differences can be as varied as the shorter cell cycle length seen in primate area 17 vs. 18 (Kornack and Rakic,1998; Lukaszewicz et al.,2005) to the earlier birth of infragranular layers seen in murine area 3 vs. 6 (Polleux et al.,1997). The factors controlling this patterning are founded and epitomized in the protomap model (Rakic,1988), yet molecular mechanisms underlying this specification remain to be unveiled. An integrated scheme based in genetic suppression as much as in enhancement can generate a complex map from a few landmark centers such as those at the cortical hem and commissural plate (Mallamaci and Stoykova,2006).

Among the many molecules important for patterning of cerebral cortex, especially in signaling, FGF family members have been on the radar essentially since the inception of the field (Mason,2007). FGF8 and related factors (consisting of FGF8, FGF17, and FGF18) are released at the commissural plate (at the anterior ridge of developing telencephalon) to control the nature and size of the rostral telencephalon. Elsewhere, at the dorsal midline, WNT and BMP proteins induce Emx2 and the development of cortical hem, choroid plexus, and hippocampus (Mallamaci and Stoykova,2006). The meticulous and comprehensive investigation of Cholfin and Rubenstein in this Journal issue takes square aim at the contributions of these sources to patterning, by examining frontal cortex (FC) subdomains in a series of cortical patterning mutants and double mutants.

This group's recent work is characterized by the remarkable potential to use a wide range of subdomain readouts, focusing on markers that have clearly defined boundaries in murine FC. The large batch of pertinent markers as well as the use of BACs in mice, showed a unique contribution of FGF17 in controlling the molecular domains of a number of these FC markers (Cholfin and Rubenstein,2007). The present study expands on the previous one by analyzing gene expression changes in embryonic telencephalon, especially in FGF17–/–;Emx2–/– double mutant. Rather expectedly, the authors find that parietal markers such as Cad6, Ephrin A5, and Lmo3 shift rostrally, medially, and dorsally in FGF8neo/neo cortex much more prominently than in FGF17–/– cortex. FGF-dependent antagonists such as Sprouty 1 and 2 (Spry) are suppressed in FGF8neo/neo much more than in FGF17–/– mutants, and the caudal transcription factors Emx2 and COUP-TFI are expanded more in the former. Oddly, rostral Ets factors and Sp8 are decreased to a greater degree in FGF17–/– mutants, a finding that could be explained by the remaining presence of MAPK antagonism (Suzuki-Hirano et al.,2005) from Spry. The authors propose that Fgf17 has a more prominent role in the local patterning of the frontal cortex, perhaps through a function of Pea3, Erm, Er81, and Sp8 as downstream targets.

The authors have also explored an important question on the specific role of Emx2 in frontal cortex development. Interestingly, they found that ventral FC in Emx2–/– mice appears to be decreased in size and is shifted dorsomedially, an effect that might be explained by an analogous shift (and increase) in the FGF-dependent Spry family, provided that Spry functions as a true antagonist of the FC transcription factors. Thus, unlike FGF17, which does not appear to regulate ventral FC development, Emx2 does have a prominent role in its patterning.

Previous studies provide evidence that FGF8 promotes expression of Foxg1 in developing cortex and suggest that Spry represses Foxg1 (Storm et al.,2003,2006). Foxg1 is important for patterning, cell cycle dynamics, and cell survival (Hanashima et al.,2002), raising the question of how it might be altered in FGF17–/– mutants. Emx2–/– mice demonstrate that a large change in Spry expression appears, and these mice have misplaced orbital FC; they provide a unique opportunity to assess endogenously (instead of electroporation-mediated overexpression as in Storm et al.,2003) whether Foxg1 area specificity conforms as expected with respect to Spry and FGF expression. Unfortunately, the authors of this study do not investigate whether Foxg1 expression is altered in Emx2–/– or FGF17–/– mutants; it could be that the FC regulation of Foxg1 is posttranslational (Regad et al.,2007) in this case.

The FGF17–/–/Emx2–/– double mutant analysis is at the heart of this paper and integrates all present work on the mutually antagonistic roles of FGFs and Emx2 in cortex patterning (Mallamaci et al.,2000; Fukuchi-Shimogori and Grove,2003; Storm et al.,2003,2006; Hamasaki et al.,2004). FGF17–/–/Emx2–/– cortex rescues the majority of domain shifts in either single mutant yet fails to rescue FGF-8 and -15 and Sp8 expansion as well as COUP-TF1 contraction seen in Emx2–/– mutants.

In summarizing this prior literature in graphic form—these pictures are worth a thousand words—Cholfin and Rubenstein enter a 5-year-old deadlock to determine whether either Emx2 or the FGFs are initially upstream of the other in cortical patterning (Fukuchi-Shimogori and Grove,2003; Hamasaki et al.,2004). Here, the authors find that Emx2 expression is nearly unchanged in FGF17–/– mice, whereas Emx2 is increased and expanded in FGF8neo/neo and in more severe Fgf8 mutants (Garel et al., 2003; Storm et al.,2003,2006). Thus, the FGF17–/– phenotype does not appear to be secondary to expanded Emx2 in this mutant, lending meaningful support to Fukuchi-Shimogori and Grove (2003). On the other hand, the modest rostral expansion of COUP-TF1 in FGF17–/– cortex could contribute to the small decrease in frontal cortex properties.

Prior work has shown that Sp8 and FGF8 exhibit reciprocal induction (Sahara et al.,2007); this correlates with the expansion of Sp8 and FGF8 in Emx2–/– cortex seen by Cholfin and Rubenstein, and neither expansion appears to be reversed in FGF17–/–/Emx2–/– double mutants. The authors purport that FGF8 is largely unaffected in FGF17–/– mutants, yet dorsal Sp8 appears quite decreased in FGF17–/– mutants, creating a potential conflict. However, the authors suggest a useful explanation. They propose that FGF expression in the rostral patterning center is organized into a “core” and “penumbra,” with FGF8 and FGF18 in the core and FGF15 and FGF17 in the penumbra. In FGF17–/– mutants, the “core” factor FGF8 maintains some mutual enhancement of Sp8 expression. However, it is clear that FGF17 is important in promoting Sp8 expression on its own in the rostral cortical primordium, especially in the “penumbra.”

Such a complex network of gene interactions raises yet another of the greatest questions following from this paper (and the preceding one): on a molecular level, how does FGF17 act differently from FGF8, and possibly in even the same cells? Current work on midbrain and cortex development hints that the answer is likely the (mathematical) product of receptor specificity and ligand concentration. In mid- and hindbrain organization, FGF signaling results in at least a tri-state of outcomes: near-zero signaling has no effect; low signaling promotes midbrain development; high signaling induces cerebellum development (Liu et al.,2003). In cortex, target cells must first express the proper receptor to mediate the right kind of signaling; FGFR3c, for example, has higher affinity for FGF8 than for FGF17 or FGF18 (Zhang et al.,2006); the level of resultant FGF signaling (i.e., high in the “core,” low in the “penumbra,” near-zero in caudolateral cortex) may then provide feedback to tailor a cell's receptor content, as it does in midbrain (Liu et al.,2003), allowing the cell to respond both appropriately and stably to later stimuli. The Spry family is likely involved throughout, as these inhibitors respond only to the “high” FGF signaling state. Similar mechanisms are inevitably at work elsewhere in area specification, where discrete regions are derived from intermediate levels of their determining factors, Lhx2 by BMPs and Foxg1 by FGFs, to name a few (O'Leary et al.,2007).

The Cholfin and Rubenstein studies also provide a hint into possible molecular mechanisms of cortical expansion and elaboration during cortical evolution. Although FGF itself might not be directly involved in this expansion, the present research shows how it could have happened at the molecular level, by establishing that the FC can be discriminated (and its size regulated) by the expression of specific transcription factors at early embryonic stages. This work extends a rich history of experimentation, starting with “anatomical engineering” using experimental manipulation of prenatal development (Rakic et al.,1991), followed by “neurocreationism” using more advanced molecular methods, such as in the seminal study from the Grove laboratory that unveiled the region-inducing properties of FGF8 (Fukuchi-Shimogori and Grove,2001; Rakic,2001). Overall, this type of research provides an unprecedented opportunity to determine not only how cortical areas develop in an individual but also how they may emerge during evolution (Rakic,1995,2001).

On the other hand, most of the research in cortical area specification has been on the major molecular determinants of cortical morphology. From Emx2, FGFs, and WNTs and BMPs to COUP-TFs, so much has been seen, yet so little is actually known. With much of the low-hanging fruit now picked, future investigations must focus on grinding out the cell- and molecule-level physiology responsible for the battling of competing regions over precious cortical real estate, as well as the enactment of area-specific phenotypes.