|Home | About | Journals | Submit | Contact Us | Français|
Cell migration is required for normal embryonic development, yet how cells navigate complex paths while integrating multiple guidance cues remains poorly understood. During brain development, interneurons migrate from the ventral ganglionic eminence to the cerebral cortex within several migratory streams. They must exit these streams to invade the cortical plate. While SDF1-signaling is necessary for normal interneuron stream migration, how they switch from tangential stream migration to invade the cortical plate is unknown. Here we demonstrate that SDF1-signaling reduces interneuron branching frequency by reducing cAMP levels via a Gi-signaling pathway using an in vitro mouse explant system, resulting in the maintenance of stream migration. Blocking SDF1-signaling, or increasing branching frequency, results in stream exit and cortical plate invasion in mouse brain slices. These data support a novel model to understand how migrating interneurons switch from tangential migration to invade the cortical plate in which reducing SDF1-signaling increases leading process branching and slows the migration rate, permitting migrating interneurons to sense cortically directed guidance cues.
Two primary pathways of migration exist for neurons destined for the cerebral cortex. Excitatory projection neurons migrate radially from the pallial ventricular zone (VZ) (Hatten, 2002;Kriegstein and Noctor, 2004). In contrast, interneurons (inhibitory neurons) have a much longer and more circuitous migration path, originating in the subpallial ganglionic eminence (GE) and migrating tangentially through the pallium, eventually turning to enter the cortical plate (Marin and Rubenstein, 2003). Interneurons predominately migrate within two defined streams; one in the marginal zone (MZ) and the second a broader stream in the sub-ventricular (SVZ) and intermediate (IZ) zones (McManus et al., 2004; Metin et al., 2006). Interneurons eventually turn out of these streams to enter the developing cortical plate, suggesting the presence of neocortical guidance factors that direct interneurons to their final position (Supplemental Fig. S1 and Lopez-Bendito et al., 2008).
Several secreted guidance factors have been identified for their role in directing interneurons along their migratory path. For example, Sema3a repels migrating interneurons from the striatum (Tamamaki et al., 2003), while Nrg1-Ig attracts interneurons to migrate dorsally from the subpallium into the pallium (Flames et al., 2004). SDF1 (stromal cell-derived factor-1; also called CXCL12) was recently identified as a molecule necessary for normal stream migration (Stumm et al., 2003). SDF1 is a chemokine that induces chemotaxis in lymphocytes, and is expressed in the cerebral cortex along the migration stream of interneurons, while the SDF1 receptor, CXCR4, is expressed on migrating interneurons (Stumm and Hollt, 2007; Stumm et al., 2007;Lopez-Bendito et al., 2008). In CXCR4 null mice, stream migration is disrupted, causing interneurons to prematurely exit the migratory streams and invade the cortical plate, resulting in abnormal interneuron distribution (Li et al., 2008; Lopez-Bendito et al., 2008). While these studies indicate that SDF1 signaling is necessary for normal interneuron distribution in the brain, it remains unknown how the loss of SDF1 signaling results in interneurons leaving their tangential trajectory and prematurely invading the cortical plate.
We and others have hypothesized that the leading process of migrating interneurons functions like an axon’s growth cone to provide guidance during cell migration (Tessier-Lavigne and Goodman, 1996; Nasrallah et al., 2006; Gopal et al., 2009; Martini et al., 2009). Interneurons have a unique migratory behavior in which the nucleus moves in a saltatory manner behind a long single or branched projection: the leading process (Metin et al., 2006; Nasrallah et al., 2006). The leading process branches repeatedly during migration, possibly to sample the surrounding environment for guidance factors (Supplemental Fig. S1D). Recent data demonstrating that interneurons preferentially branch towards attractants such as Nrg1 support this hypothesis (Martini et al., 2009). In this study we test the hypothesis that SDF1 signaling influences interneuron branching in order to control interneuron stream migration and stream exit. We find that SDF1 suppresses branching, thus keeping migrating interneurons within their migratory stream and moving rapidly. When SDF1 signaling is reduced, interneuron branching is stimulated, the rate of migration is slowed, and interneurons more frequently turn and exit their migratory streams.
CD1 mice or CD1 mice crossed to Dlx5/6-Cre-IRES-eGFP mice (Stenman et al., 2003) to genetically label interneurons derived from the medial ganglionic eminence (MGE) with green fluorescent protein (GFP) were used for all experiments. All studies were approved by the institutional animal care and use committee.
Brains from embryonic day 14.5 (E14.5) mouse embryos were dissected in ice-cold HBSS (Sigma), embedded in 4% low melting point agarose (Lonza) in HBSS and sectioned coronally at 250 um. Slices were placed on Millicell-CM culture inserts (Millipore), while 100 um square explants were cut out of the MGE of the slice and placed on plastic dishes or glass coverslips. The explants were covered with 50% Matrigel, 50% Collagen (2 mg/ml, BD Biosciences), and placed at 37°C, 5% CO2 for 20 minutes to gel. Tissue was incubated in DFS media (F12:DMEM w/ 10% FBS, 40 uM L-glutamine, 47 mM glucose, P/S) for one hour, then switched to DM media (DMEM w/N2, 36 mM glucose, P/S). Slices were imaged after one hour in DM media, while explants were imaged after one day.
Speed and branching analyses were performed on time-lapse images acquired at 10X magnification and at 10 min intervals for 8–12 hours on a Nikon TE300 microscope equipped with an Okolab environmental chamber 37°C, 5% CO2. Time-lapse imaging of GFP labeled interneurons migrating in brain slices were acquired with 5 um Z-stacks at 20X at 10 min intervals for 3.5 hours on a Leica SP2 or Olympus FV10i confocal microscope. SDF1-bead-implanted slices were imaged on a Leica MZ16FA stereomicroscope.
Interneurons were exposed to SDF1a by co-culturing COS7 cell clumps co-expressing SDF1a and dsRed 300 um from MGE explants within Matrigel:Collagen (BD Biosciences). COS7 cells were transiently transfected with Fugene 6 (Roche) 2 days prior to pelleting and resuspension in an equal volume of Matrigel:Collagen to form clumps. To inhibit Gi signaling, slices were electroporated with a construct expressing interfering peptide IKNNLKDCGLF (cue BIOtech), followed by setting up explants as described above. Modulation of cAMP was achieved by treating tissue with 125 uM forskolin (Sigma) or 790 uM db-cAMP (Sigma) and imaged immediately. To block SDF1 signaling, slices were treated with the CXCR4-inhibitor AMD3100 (Sigma, 189.6 uM). To treat slices with BSA or recombinant SDF1 (PreproTech) agarose beads (Affi-Gel, Bio-Rad) were soaked in 20ug/ml protein solution 1–24 hours at 4C and imbedded into the slice at defined positions.
Interneuron average speed was measured by recording cell soma position in each frame using the Manual Tracking plugin in ImageJ (NIH, http://rsb.info.nih.gov/ij/) and averaging the speed from each frame in the interval of analysis. Interneuron branching was recorded manually. Interneurons with blocked Gi signaling were identified by the fluorescence of co-transfected dsRed. Unfortunately, the fluorescence was not bright enough to allow reliable quantification of branching; speed was analyzed instead. Interneuron distribution across the neocortex was determined by manual counting using the Cell Counter plugin in ImageJ. Cells were counted in a defined 400 pixel (293 um) wide area of analysis divided into 10 equal bins, placed 300 um dorsal to the pallial/subpallial boundary.
Statistical analysis was carried out using Microsoft Excel or R (www.r-project.org) with plugin R-commander. If data appeared normally distributed, we asked whether there were differences between groups using multi-way ANOVA and with post-hoc tests adjusted for multiple comparisons using a Bonferroni correction. For forskolin and db-cAMP treatment in vitro, the branching data appeared non-normally distributed, so a similar approach was used, but in this case a generalized linear model with a Poisson distribution and a term to correct for overdispersion was used to fit the data, yielding an overall test of significance based on Analysis of Deviance. Interneuron distribution in AMD3100- and forskolin-treated slices and branch angle distributions were analyzed using a Chi-squared test. If a difference was found between the populations, control and treated values for each bin were independently analyzed using two by two contingency tables.
SDF1 signaling is necessary for normal interneuron stream migration, although SDF1’s affect on interneuron migratory behavior is incompletely understood. In contrast, Nrg1 is known to alter interneuron branching to attract interneurons towards the pallium. We have hypothesized that SDF1 also influences branching behavior in order to facilitate interneuron migration within streams.
We examined the branching behavior of control- and SDF1-treated interneurons in vitro by timelapse microscopy (explant culture). Interneurons treated with SDF1 migrate faster than controls, as previously reported (Lopez-Bendito et al., 2008), but were also found to branch less frequently (Fig. 1). The SDF1-induced branching phenotype was found for both leading process splitting as well as interstitial branching (Fig. 1D). These data demonstrate that SDF1-signaling enhances the rate of migration while simultaneously affecting specific branching behavior; a reduction in branching frequency.
In vivo, stream migration occurs in areas of SDF1 expression (Stumm et al., 2007; Lopez-Bendito et al., 2008), thus we hypothesized that interneurons migrating within SDF1-rich streams into the pallium would exhibit reduced branching and fast migratory rates. As these neurons move from ventral to dorsal in the SVZ/IZ stream, they are exposed to progressively lower SDF1 levels (Fig. 2A); thus we predicted they would exhibit increased branching and slower migration. This would lead to increased exploration of their environment, increased directional changes, and ultimately accelerated cortical plate invasion.
To test this model in brain slices, we measured interneuron speed and branching frequency within the SVZ/IZ stream in two positions of differing SDF1 expression by timelapse microscopy (Fig. 2B). The first position was just above the cortical notch where SDF1 is highly expressed (ventrolateral neocortex). Interneurons were found to branch at a modest rate (0.24 branches/hr) and migrate quickly (53 um/hr, Fig. 2C,D). In contrast, farther up the cortical arch where SDF1 expression is lower and narrower (dorsomedial neocortex), interneurons branch about 2.5 times more frequently (0.6 branches/hr) and slow down (47 um/hour). We also measured the angle between leading process branches as differences in branch angle have been previously observed at different areas along the interneuron migration path (Martini et al., 2009). Interneurons migrating in the SVZ/IZ stream near the notch had an average branch angle of 50±2.2°, while interneurons migrating farther along in the arch had a slightly higher average branch angle of 56±2.2°. The distribution of branch angles in both areas appeared log-normal but only in the arch area did branch angles larger than 105° occur, with significantly wider angles occurring throughout the distribution (Fig. 2E,F).
To test the pathfinding aspect of our model, we acutely inhibited interneuron SDF1-signaling in brain slices. As expected, inhibition of SDF1-signaling with the CXCR4 inhibitor AMD3100 disrupted the migration stream organization, as seen previously, and similar to the CXCR4 null mouse phenotype (Fig. 3A,B) (Lopez-Bendito et al., 2008). Importantly, inhibiting SDF1-signaling increased the branching frequency of interneuron leading processes in the SVZ/IZ migration stream (Fig. 3C) and resulted in a shift in interneuron distribution. Interneurons in control slices were most frequently present in two main migration streams (SVZ/IZ and MZ), while interneurons in AMD3100-treated slices exited out of the SVZ/IZ migratory stream to find positions in the cortical plate and VZ (Fig. 3D). Thus, we find SDF1 signaling reduces branching frequency of migrating interneurons both in explant cultures and in brain slices. These data suggest that SDF1 induces a specific branching behavior in interneurons that is necessary for normal stream migration and interneuron position within the developing cortex.
The SDF1 signaling pathway governing this branching behavior has yet to be investigated in migrating interneurons, but SDF1-signaling is known to positively modulate leukocyte chemotaxis and primordial germ cell migration in zebrafish (Phillips and Ager, 2002; Knaut et al., 2003). The SDF1 receptor CXCR4 functions as a G-coupled protein receptor that can act through Gi to inhibit adenylyl cyclase (AC), lowering cyclic adenosine monophosphate (cAMP) levels and lowering protein kinase A (PKA) activity. Given that other migrating cell types employ similar signaling mechanisms during migration, we predicted that the increased speed of migration and reduction in branching frequency seen in SDF1-treated interneurons in vitro would be the result of a similar signaling pathway.
To test this we electroporated interneurons in vitro with a cueBIOtech Gi minigene vector expressing an interfering peptide known to specifically block Gi signaling (explant culture). The Gi minigene had little effect on migration in the absence of SDF1, but in the presence of SDF1, blocking Gi signaling blocked the expected increase in migration rate (Fig. 4A).
Given that SDF1-signaling normally inhibits adenylyl cyclase (AC) through Gi, we expected stimulation of AC to block the reduction in branching frequency seen in SDF1-treated interneurons. Stimulation of AC with forskolin increased branching frequency in control treated neurons and blocked reduction of branching in SDF1-treated interneurons (Fig. 4B). Similarly, stimulating protein kinase A with di-bromo cAMP also resulted in an increased branching frequency, but in the presence of SDF1, di-bromo cAMP blocks the expected reduction in branching (Fig. 4C). These data indicate that SDF1 inhibits branching of interneurons through this Gi-coupled receptor signal transduction pathway.
To further test the sufficiency of increased branching to induce the same shift in interneuron distribution seen when inhibiting SDF1-signaling we again employed our slice culture system. We acutely increased the frequency of interneuron branching in brain slices by adding forskolin as was done in explant culture. This resulted in a similar disruption of the migration streams as seen in brain slices with inhibited SDF1-signaling (compare Fig. 5A,B to Fig. 3A,B). As expected, this treatment increased the branching frequency of interneurons and resulted in a shift in interneuron distribution (Fig. 5C,D). Interneurons exited from the SVZ/IZ migration stream to positions in the cortical plate. Importantly, treatment with forskolin did not effect interneurons’ capacity to respond to SDF1 in brain slices in a bead attraction assay (Supplemental Fig S2), supporting our model that either inhibiting SDF1-signaling or increasing branching frequency is sufficient to allow interneurons to exit at least one of their migration streams and invade the cortical plate.
Interneurons in CXCR4 null mice prematurely exit their migration streams and migrate to the cortical plate, eventually exhibiting an abnormal distribution (Li et al., 2008; Lopez-Bendito et al., 2008). Our data indicate that SDF1 increases the speed of migration while decreasing the frequency of branching. Based on these findings, we predicted that branching and speed would be co-regulated by cAMP. To test this hypothesis, we treated interneurons with different concentrations of forskolin to induce branching and measured interneuron migration speed. We found that as branching frequency increases, migration speed decreases (Fig. 6). At the highest concentrations of forskolin, we observed cells branching most frequently, while some cells also frequently change the polarity of their leading process and migrate quickly back and forth, resulting in a higher average speed. Cells treated with lower doses of forskolin branch more frequently than control cells while maintaining the polarity of their leading process, resulting in the lowest migration speed. These data support our model that intracellular signaling pathways that affect branching frequency through cAMP levels or PKA activity inversely affect speed.
Understanding how migrating neurons navigate the complex environment of the embryonic brain is a fundamental question in central nervous system development. Although progress has been made in understanding how migrating cells move and change their direction in vitro, the mechanisms migrating cells employ as they are directed through tissue by multiple guidance factors are largely unknown.
We have examined interneuron stream migration in the cortex and tested SDF1’s role in stream maintenance and interneuron stream exit. Although SDF1 had previously been identified as a modulator of migration rate in vitro, data we were able to confirm, we also determined that SDF1 signaling reduces interneuron branching frequency both in vitro and in brain slices. In this context, inhibiting SDF1 signaling results in interneurons prematurely exiting from migratory streams and shifting the interneuron distribution from the SVZ/IZ to the cortical plate. We found that inducing branching in migrating interneurons is sufficient to cause cells to exit their migration stream and invade the cortical plate in brain slices. While treatment with AMD3100 will act autonomously on interneurons, we have not yet excluded the possibility that treatment with forskolin could act non-autonomously to cause CP invasion. Taken together our data suggest that SDF1 signaling suppresses branching frequency to allow cells to maintain a simpler morphology better suited to rapid migration within a defined stream. Interestingly, these treatments did not cause a measurable amount of interneurons to exit the MZ stream where SDF1 is also highly, but more uniformly expressed (Fig. 3,,5,5, bin 10). This could suggest intrinsic differences between the SDF1 response of interneurons occupying the two different migration streams, external substrate differences, or a more complex guidance switch with additional signaling factors to induce interneurons to exit the MZ to invade the cortical plate.
To characterize the intracellular signaling pathway used by SDF1 in interneurons, we interrogated known components of the pathway. We expected CXCR4 signaling to occur through Gi as in other cell types. We found that Gi regulation of cAMP concentration is important in defining interneuron speed and branching frequency. We also found that protein kinase A may transduce this 2nd messenger signal to further signaling pathways to create the complex migration behavior of interneurons.
Our data imply an inverse relationship between branching and the rate of migration. Treating interneurons with SDF1 resulted in both an increase in speed and a decrease in leading process branching. We next asked if regulation of these two aspects of cell behavior diverge farther down the SDF1 signaling pathway. Treatment over a range of forskolin concentrations, which increases cAMP through adenylyl cyclase sensitization, determined that increasing cAMP concentration increases branching while reducing speed. Taken together with our SDF1 treatment results, cAMP concentration is a central mediator of both speed and branching frequency. Whether cAMP affects speed and branching through separate downstream signaling pathways or through shared cytoskeletal regulation remains to be determined.
The inverse relationship between branching and speed has led us to propose a new model to describe interneuron stream migration (Fig. 6B). In this model, interneurons migrating in streams, under the influence of SDF1, migrate faster as a result of a reduced branching frequency. This allows interneurons to spread from their ganglionic eminence origins through the cortex and achieve a proper interneuron distribution. As individual interneurons encounter lower SDF1 levels, and consequently reduce SDF1 signaling, they increase their branching and reduce their speed. This permits them to sample as yet unknown cues, change direction and migrate into the cortical plate to achieve their final position.
One possible cortical plate invasion cue could be a time-dependent reduction in SDF1-signaling levels. It is likely that SDF1 signaling within individual interneurons is reduced as they exit migration streams (Liapi et al., 2008; Lopez-Bendito et al., 2008). Consistent with our observation of increased branching and stream exit when treating brain slices with SDF1-signaling inhibitor, migratory interneurons in the CXCR4 knock out mouse exit their migration streams prematurely, demonstrating that a reduction in SDF1-signaling allows cells to exit migration streams but also that SDF1 alone cannot define and maintain stream migration (Li et al., 2008). Whether this reduced SDF1-signaling is the result of downregulation of CXCR4 or internally initiated interference with the SDF1-signaling pathway, perhaps at the cAMP level, remains to be elucidated. One external cue for interneuron-stream exit is likely decreasing SDF1 concentration as interneurons migrate down the gradient of SDF1 expression (Stumm et al., 2007). Consistent with this hypothesis, our measurement of interneuron branching in brain slices shows that interneuron branching is lower near the cortical notch, an area of high SDF expression, but farther up the cortical arch where SDF expression is reduced, interneuron branching increases (Fig. 2).
According to our data and model, as SDF1 signaling is reduced, the branching frequency of interneurons increases (Fig. 6B). We speculate that this increase in branching allows interneurons to probe greater space and come into contact with additional guidance factors. A cortical plate attractant has been biologically demonstrated, but is yet to be identified (Lopez-Bendito et al., 2008). In addition to the cortical plate attractant, it is possible that repellents could direct interneuron migration in the cortex. There is some evidence that semaphorins may be expressed in the dorsal VZ/SVZ (Tamamaki et al., 2003) and SDF1 has been shown to reduce the response of pathfinding axons to repellents in the Slit and Semaphorin families (Chalasani et al., 2003). In the cortex, interneurons migrating in SDF1-rich streams would be insensitive to any repellents. Upon reduction in SDF1-signaling, interneurons could then respond to the repellents and exit migration streams toward the cortical plate. Interestingly, we provide some evidence that reduction of SDF1 signaling with AMD3100 can slightly increase interneuron localization in the VZ (Fig. 3, bin 1), suggesting that blocking SDF1 signaling may leave interneurons free to explore the VZ. In contrast, if SDF1 signaling is intact but interneuron branching increased by forskolin treatment, fewer cells localize to the VZ, even though cells exit the SVZ/IZ stream on both sides, possibly displaying a balance between SDF1’s attractive ability and interneuron branching causing more exploratory migration (Fig. 5, bins 1 and 4,5,6). Finally, in our model increased branching would shift the balance of stimulation from SDF1 to the cortical plate attractant or a VZ localized repellent, resulting in slower migration and stream exit. The role of SDF1 in this model is to affect pathfinding by modulating cell morphology, not necessarily for SDF1 to function as an attractant or a motogen, both additional possible functions (Li et al., 2008; Lopez-Bendito et al., 2008).
Our data provides a new model as to how SDF1 controls interneuron stream migration through signal transduction and modification of cell migratory morphology and behavior. Because we have shown that changes in branching frequency are relevant in a tissue context, we expect future studies to further define the role of migrating interneuron branching in brain development.
Fig1video1-2. SDF1 reduces branching frequency and increases speed. Primary interneurons migrating in vitro from a medial ganglionic eminence explant treated with control (video1) display leading process and interstitial branching. Interneurons treated with SDF1 display increased migration speed and reduced frequency of branching (video2). 10X time-lapse phase microscopy, 1 frame / 10 min, 3 FPS.
Fig3video1-2. Interneurons migrate in defined streams in Dlx5/6-Cre-IRES-GFP mouse brain slices (video1). AMD3100 disrupts interneuron stream migration by increasing branching frequency in slices and results in increased cortical plate invasion (video2). 20X time-lapse confocal microscopy, 1 frame / 10 min, 7 FPS.
Fig4video1-3. Control-treated DsRed-transfected primary interneurons migrate in vitro at typical speed, while SDF1-treated interneurons migrate faster (video2). Co-transfection with Gi-minigene does not significantly change migration speed in control-treated interneurons (video1), but in SDF1-treated interneurons, co-transfection with the Gi-minigene returns migration speed to control levels (video3). 10X time-lapse epifluorescence microscopy, 1 frame / 10 min, 3 FPS.
Fig4video4-5. Control-treated primary interneurons display basal branching frequency in vitro, while treatment with forskolin increases branching significantly (video4). Treatment with db-cAMP results in a similar increase in branching frequency (video5). Both these treatments block SDF1-induced reduction in branching frequency. 10X time-lapse phase microscopy, 1 frame / 10 min, 3 FPS.
Fig5video1. Forskolin-treatment disrupts interneuron stream migration by increasing branching frequency in mouse brain slices and results in increased cortical plate invasion, similar to blocking SDF1-signaling with AMD3100. 20X time-lapse confocal microscopy, 1 frame / 10 min, 7 FPS.
We would like to thank the members of the Golden Lab for their helpful comments and support. We thank Dr. Jonathen Raper for the kind gifts of SDF1a and Gi inhibitory expression vectors. We thank Gaia Colasante and Dr. Vania Broccoli for the kind gift of an SDF1a vector. We thank Dr. Ramon Pla, Juan Antonio Sanchez, and Dr. Oscar Marin for assistance with statistical analysis. We thank Drs. Jonathan Raper and Greg Bashaw for critically reviewing this manuscript. This work was supported by National Institutes of Health grants NS45034 and HD26979.