Translation of the cell adhesion molecule ALCAM in axonal growth cones – regulation and functional importance

During development of the nervous system, cell adhesion molecules of the immunoglobulin superfamily (IgSF-CAMs) are crucial for neuronal migration, axon elongation and navigation (Maness and Schachner, 2007). The IgSF-CAM ALCAM (for activated leukocyte cell adhesion molecule, previously also termed DM-GRASP, SC1, BEN, CD166 or MEMD) plays a role in cell adhesion (Burns et al., 1991; Tanaka et al., 1991), neuronal migration and differentiation (Stephan et al., 1999; Heffron and Golden, 2000), axon growth (Pollerberg and Mack, 1994; DeBernardo and Chang, 1995) and axonal pathfinding (Avci et al., 2004; Weiner et al., 2004). ALCAM interacts homophilically, i.e. with itself (Tanaka et al., 1991; van Kempen et al., 2001), and heterophilically with the IgSF-CAM CAML1 (also known as NgCAM) (DeBernardo and Chang, 1996; Buhusi et al., 2009). ALCAM is an integral membrane protein, which is ubiquitinated, rapidly endocytosed [in a manner dependent on clathrin or extracellular-signal-regulated kinase (ERK)] and delivered to the degradation pathway; this does not cause, however, changes in the level of ALCAM in the growth cone plasma membrane (Thelen et al., 2008).

ALCAM is present in extending and fasciculating axons and is absent from their somata. In the early developing retina, for example, ALCAM is only present on growing retinal ganglion cell (RGC) axons (Pourquie et al., 1992; Pollerberg and Mack, 1994). RGC axons extend while contacting pre-existing RGC axons (in the optic fiber layer of the retina), which results in bundle formation (fasciculation). In assays mimicking this scenario in vitro, RGC axons prefer to grow on ALCAM-containing lanes, indicating that these axons also prefer ALCAM-providing structures in vivo (Avci et al., 2004). When ALCAM is functionally neutralized by antibody inhibition or knocked out, RGC axon fasciculation and entry into the optic nerve is reduced (Avci et al., 2004; Weiner et al., 2004). Inhibition of ALCAM interactions (Avci et al., 2004) or of its internalization (Thelen et al., 2008) abolishes the axonal preference for ALCAM-containing lanes, demonstrating the impact of physiological levels of ALCAM in the plasma membrane. Highly defined nano-spaced ALCAM, offered as cell culture substrate, revealed the importance of a critical density of this CAM for axon growth (Thelen et al., 2007; Jaehrling et al., 2009).

ALCAM does not possess any sorting signals targeting the protein to the axonal compartment, such as found for CAML1 (Kamiguchi and Lemmon, 1998; Yap et al., 2008); we therefore asked whether ALCAM might be synthesized in the axon tip (growth cone). Translation in the axonal growth cone has been shown for approximately ten different proteins (Jung and Holt, 2011) and is required for the response to various guidance cues (Lin and Holt, 2007). Two RNA-binding proteins, cytoplasmic polyadenylation element-binding protein (CPEB1) (Huang et al., 2003) and zipcode-binding protein (ZBP1) (Leung et al., 2006; Yao et al., 2006), have been identified as mediating the transport of mRNA and its translation in the growth cone by binding to kinesins and the 3′-untranslated region (3′-UTR) of mRNAs (Lin and Holt, 2007). Here, we show that ALCAM is translated in growth cones. This translation is 3′-UTR-regulated, depends on the kinases ERK and TOR, prevents overexpression and compensates for the endocytotic degradation of ALCAM, together balancing the physiological density of this CAM in the growth cone plasma membrane. The local translation of ALCAM enhances RGC axon growth and is crucial for the preference of RGC axons for ALCAM-providing structures, indicating the importance of its growth cone translation for axonal elongation and navigation in the developing nervous system.

In the developing retina, ALCAM protein is selectively present on RGC axons (Fig. 1A). The axons of the RGCs are the only retinal axons that grow towards the optic nerve head [i.e. from the periphery to the center; at embryonic day (E) 4–8], dive into the optic nerve and ultimately project to the mesencephalon. Because these axons have to cover a considerable distance and display ALCAM far away from the soma [at the optic chiasm, ~6 mm away at E9 (Pollerberg and Mack, 1994)], we asked whether ALCAM might be produced not only in the RGC somata but also in their growth cones. We therefore visualized the distribution of ALCAM mRNA in the embryonic retina by in situ hybridization (Fig. 1B,C): ALCAM mRNA was present in the entire retina, including the ganglion cell layer (formed by the somata of the RGCs), however, not at elevated levels. This points to another production site for ALCAM protein in addition to the somatic one, and ALCAM mRNA is indeed present in the optic fiber layer, which is composed solely of RGC axons and their growth cones. ALCAM mRNA is also found in a subpopulation of immature neuroepithelial cells, which do not possess detectable levels of ALCAM protein, indicating that the translation of ALCAM is suppressed in these cells. ALCAM protein is present in the plasma membrane of RGC axons and growth cones extending in retinal cell culture (Fig. 1D) and fluorescence in situ hybridization (FISH) experiments revealed the presence of ALCAM mRNA in these axons and growth cones (Fig. 1F,G). We separated the distal region of RGC axons extending from retinal explant strips (which we called the ‘growth cone and distal axon fraction’); the remaining proximal RGC axon parts and adjacent explants, containing the somata, were also harvested (the ‘explant fraction’) (Fig. 1E). Both fractions were tested for presence of ALCAM mRNA by RT-PCR. Histone-H3-encoding mRNA (which is selectively present in somata) and β-actin-encoding mRNA (known to be translated in growth cones) were also analyzed; the absence of histone mRNA from the growth cone and distal axon fraction demonstrates that this isolate does not contain somatic mRNA. By contrast, ALCAM mRNA is clearly present in the growth cone and distal axon fraction as well as in the explant fraction.

To gain direct evidence that ALCAM protein is produced in growth cones, single RGC axons, extending from retinal explant strips, were cut in the distal region, separating the growth cones from their proximal axons and somata. The disconnected growth cones continued to advance and 6 hours after the cut, they were fixed and immunofluorescence-labeled to visualize ALCAM protein (Fig. 2). Fluorescence intensity quantification showed that the ALCAM level in the plasma membrane of separated growth cones was not significantly changed compared with that in growth cones of uncut RGC axons. Double labeling with another IgSF-CAM, CAML1, which is known to be produced in the soma and transported to the growth cone, was performed for comparison. CAML1 levels in separated growth cones decreased to <60% of the values in uncut growth cones. To find out whether local protein synthesis is the underlying mechanism for the maintenance of the ALCAM level in isolated growth cones, translation was inhibited by addition of cycloheximide. This treatment caused a clear decrease in ALCAM levels in the plasma membrane of the separated growth cones (35%), whereas the level of CAML1 did not significantly change. Taken together, these data show that the level of ALCAM protein in the plasma membrane of growth cones largely depends on local translation in this structure.

The amount of ALCAM in the growth cone plasma membrane depends not only on synthesis but also on the degradation of ALCAM, and is crucial for proper growth cone functioning (Thelen et al., 2008); ALCAM is mainly present in the axonal plasma membrane and is also enriched at the growth cone in vitro (supplementary material Fig. S1). We therefore examined the impact of the translation of ALCAM on the presence of ALCAM in the growth cone plasma membrane at the same as inducing the internalization of ALCAM. For this, translation was inhibited by the TOR inhibitor rapamycin in retinal single cell culture and endocytosis of ALCAM was triggered by clustering of cell surface ALCAM using antibodies (Fig. 3). A clear reduction of ALCAM was already detected by 10 minutes after the addition of rapamycin (−41±6%; P<10⁻⁷), whereas in controls (without translation inhibition), ALCAM levels in the growth cone plasma membrane were not significantly decreased (−16±4%; P<0.12). After 60 minutes, the induced internalization had further decreased ALCAM levels in growth cones with or without translation inhibition (−74%, P<10⁻⁴ and −59%, P<0.001, respectively), however, it had decreased to a clearly lesser degree in growth cones where translation is ongoing (−35%; P<10⁻⁵). Taken together, these data show that the translation of ALCAM in the growth cone is important for a rapid and persistent compensation of the removal of plasma membrane ALCAM due to endocytosis.

Because we could show both the translation of ALCAM and the presence of its mRNA in growth cones, we sequenced the flanking regions of ALCAM mRNA isolated from (E6) retina and analyzed them for mRNA transport and/or translation regulating sequences (supplementary material Fig. S2). This revealed that the 3′-UTR of ALCAM contains cytoplasmic polyadenylation elements (CPEs), conserved regulatory elements that are known to contribute to transport and local translation of mRNAs. To examine what role the 3′-UTR plays in ALCAM translation in growth cones, we fused its 3′-UTR to a membrane-anchored, rapidly destabilized GFP reporter (dGFP), which – when somatically produced – does not contribute to the fluorescence signal measured in the growth cone, and transfected the construct into retinal single cell cultures. The 3′-UTR of γ-actin-endoding mRNA, which is known to be exclusively translated in the soma, was used as a control. The reporter fluorescence intensity in growth cone and axon (the medial region and region proximal to the soma) was determined (Fig. 4). In RGCs transfected with dGFP fused to the 3′-UTR of γ-actin-endoding mRNA, the dGFP signal in the growth cone was low [−8.5% compared with that in the proximal axon (no significant difference), n=51, P<0.65], indicating that no detectable levels of fluorescence for a somatically produced reporter reach the growth cone. By contrast, axons of RGCs transfected with dGFP fused to the ALCAM 3′-UTR, exhibited a strong dGFP signal in the growth cone (+120.8% compared with the proximal axon: n=28, P<0.02), demonstrating ALCAM 3′-UTR-mediated translation in the growth cone.

To confirm further the translation of ALCAM in growth cones, we fused the 3′-UTR of ALCAM to a reporter gene encoding a fluorescent protein which turns from green to red with time (pTimer) so that no green fluorescent signal arrives in the growth cone from somatically produced reporter. The construct was transfected into retinal single cell cultures, and levels of young (green) and aged (red) reporter were determined in RGC somata, axons and growth cones. Newly translated reporter was clearly present in growth cones; compared with the medial axon, a growth cone displayed a more than twofold increase in fluorescence intensity (Fig. 5A,B). Deletion of the ALCAM 3′-UTR caused a clear reduction (of absolute, as well as relative, intensity, Fig. 5B,C) of young reporter in the growth cone compared with that in the undeleted construct. By contrast, levels in soma and axon were not significantly changed. To analyze selectively the role of the CPEs in ALCAM, we blocked their function by antisense oligonucleotides covering the CPEs and their ALCAM-specific flanking regions in retinal cell cultures expressing the reporter pTimer (Fig. 5D). Analysis of the young reporter fraction revealed that blockage of the CPEs caused a clear decrease in the relative fluorescence intensity in the growth cone and distal axon (compared with the sense-treated controls), whereas levels were unchanged in all other regions of the RGC. We also transfected a construct containing mutations in the CPEs [known to eliminate their function (Brittis et al., 2002)] in retinal single cell cultures (Fig. 5E) resulting in an even stronger decrease of young reporter in growth cone and distal axon (compared with the non-mutated 3′-UTR). The young reporter fraction was also decreased in the medial axon; the young reporter fraction was not, however, lowered in the proximal axon and soma, showing that the somatic translation of ALCAM is CPE-independent. Taken together, these data show that ALCAM is translated in growth cones and that this local translation depends on its intact 3′-UTR.

Because protein expression in growth cones is regulated by intracellular signaling, we investigated the impact of the two key intracellular signal proteins TOR and ERK, which are known to activate growth cone translation upon stimulation by external cues (supplementary material Fig. S3A,B). Inhibition (by rapamycin) of TOR, a kinase which is crucial for translation initiation, in retinal cell cultures led to a decrease in the young reporter fraction in the RGC growth cone (−23%; n=19, P<0.0002), distal axon (−18%; P<0.02) and medial axon (−22%; P<0.0003) relative to that in uninhibited RGCs. Inhibition (by U0126 through MEK) of ERK, a kinase known to support translation when activated by plasma membrane receptors, causes a decrease in the young reporter fraction in the growth cone (−21%; n=26, P<0.0002) and distal axon (−21%; P<0.0005) but not, however, in the medial axon. This stronger restriction to the axon tip points to a more local role of ERK for translation regulation of ALCAM in the growth cone than that of TOR.

We also assessed the question of whether blockage of ALCAM translation by oligonucleotides directed against its 3′-UTR CPEs affects the presence of ALCAM protein in the growth cone plasma membrane. For this, cell surface ALCAM was labeled and fluorescence intensities in soma, the proximal, medial and distal axon, and growth cone were quantified in retinal cell cultures treated with antisense or sense oligonucleotides. The presence of ALCAM in the growth cone plasma membrane was clearly reduced by the antisense oligonucleotides compared with controls (−19.4%; n=68 and n=71, respectively; P<0.03), whereas levels in the axon region proximal to the soma were not significantly changed (−11.1%; P<0.2). This indicates that the somatic production of ALCAM is unaffected and points to a 3′-UTR independent translation in the soma (as also indicated by the pTimer experiments, see above). These findings show that the 3′-UTR-CPE-regulated translation of ALCAM in the growth cone contributes towards maintaining physiological ALCAM levels in the plasma membrane of this cellular compartment.

We next tested whether the 3′-UTR of ALCAM is capable of limiting the level of ALCAM protein in the growth cone plasma membrane. For this, retinal cell cultures were transfected with ALCAM cDNA, including the complete 3′-UTR, and ALCAM protein levels in the plasma membrane were quantified. ALCAM cell surface levels of RGCs, however, did not increase compared with those in untransfected RGCs (Fig. 6A; supplementary material Fig. S1). Transfections at an earlier stage, i.e. before onset of endogenous ALCAM expression in the eye vesicles (in intact E1.5 embryos kept for in 3 days in ovo) also do not result in the expression of detectable levels of ALCAM in the retina (Fig. 6B). Moreover, also cerebellar neurons, which express ALCAM endogenously as do RGCs, do not display higher ALCAM levels in the plasma membrane after transfection (supplementary material Fig. S4A). By contrast, N2A and HEK293 cells, which do not express ALCAM endogenously, exhibit high levels of cell surface ALCAM upon transfection (supplementary material Fig. S4B,C). Only when ALCAM cDNA containing the 3′-UTR of the bovine growth hormone (as a control) was transfected into retinal cell culture, were ALCAM levels in the plasma membrane of RGCs are strongly elevated (Fig. 6A; supplementary material Fig. S5). Taken together, these findings show that the 3′-UTR of ALCAM limits the expression of ALCAM to physiological levels and that this control is only active and/or present in endogenously ALCAM-expressing cells.

To examine the functional relevance of physiological ALCAM levels in the growth cone plasma membrane, we reduced its expression in RGC axons that grow on a substrate of either laminin-coated coverslips or on coverslips coated with laminin and ALCAM. The latter substrate allows for trans-interactions between ALCAM in the axonal plasma membrane and ALCAM offered by the substrate, as occur between RGC axons extending on axons in vivo. RGC axon growth was enhanced (+17%) on the ALCAM plus laminin substrate compared with the growth on laminin only (average length: 137±7 μm, n=96, and 117±6 μm, n=99, respectively; P<0.03) (Fig. 7A). When ALCAM translation was reduced by oligonucleotides directed against its CPEs (decreasing ALCAM levels in the growth cone), RGC axon extension on the ALCAM plus laminin substrate was decreased (−17%) compared with that in controls treated with sense oligonucleotides (117±6 μm, n=99, and 136 ±8 μm, n=96, respectively; P<0.012), thereby (exactly) neutralizing the enhancement of axon growth induced by ALCAM present in the substrate. On laminin substrate, by contrast, this oligonucleotide treatment has no effect on axon growth. Reduction of growth cone translation of ALCAM (by ERK inhibition through U0126) reduces axon growth (compared with that in controls treated with DMSO) was in a similar range (−21%; 99±6 μm, n=174, and 126±9 μm, n=196, respectively; P<0.03) to that upon treatment with the antisense oligonucleotides. Taken together, the data obtained by blocking ALCAM translation in two ways show that physiological ALCAM levels in the growth cone plasma membrane are a pre-requisite for ALCAM-induced enhancement of RGC axon growth.

We also knocked down ALCAM expression by shRNA transfection in dense (E6) retinal cell cultures where RGCs send out their axons in contact with other RGC axons, and thus ALCAM–ALCAM trans-interactions are taking place as in the (E6) developing retina where RGC axons also extend on each other. Axon elongation of RGCs transfected with ALCAM shRNA is clearly reduced (−61%) compared with that in non-transfected RGCs (axon length: 69±5 μm, n=108, and 172±7 μm, n=102, respectively; P<1.2×10⁻²⁴) or control-shRNA-transfected RGCs (173±10 μm, n=107; P<1.8×10⁻¹⁶) (Fig. 7B; supplementary material Fig. S6). The stronger reduction of axon growth due to the global ALCAM knockdown than the one caused by inhibition of the local growth cone translation of ALCAM (by ERK inhibitor or antisense oligonucleotides) suggests that there also is a 3′-UTR-independent production mechanism for ALCAM. In addition, overexpression of ALCAM (achieved by transfection of ALCAM cDNA containing the 3′-UTR of the bovine growth hormone) results in a reduction (−31%) of RGC axon growth compared with that in mock-transfected RGCs (67±5 μm, n=92, and 93±8 μm, n=85, respectively; P<0.006) (Fig. 7C). Hence, both manipulations of ALCAM translation – whether causing decrease or increase in its plasma membrane density – negatively affect RGC axon growth, indicating that only a physiological ALCAM level ensures efficient axon advance.

We also addressed the question of whether manipulations of the ALCAM level in the growth cone plasma membrane have an impact on the known preference of RGC axons for ALCAM-containing substrates (Avci et al., 2004). For this, we employed an assay where alternating substrate lanes present ALCAM plus laminin (as in the optic fiber layer formed by RGC axons) or laminin only (as in the rest of the retina) to RGC axons extending from retinal explant strips. Under control conditions, almost all RGC axons grew on the ALCAM-containing lanes and avoided the ALCAM-free lanes (this preference was observed in 83% of the explants; 221 boundaries were evaluated, n=12) (Fig. 8A). Reduction of ALCAM growth cone translation by ERK inhibition causes a complete loss (preference 0%, 333 boundaries evaluated, n=14) of axon preference for ALCAM-presenting lanes, as RGC axons readily trespass onto the ALCAM-devoid Laminin lanes (Fig. 8B). When ALCAM expression is decreased by antisense oligonucleotides directed against its 3′-UTR, only a weak axonal preference (25%, 323 boundaries were evaluated, n=12) for ALCAM lanes was observed. By contrast, in all cultures treated with sense oligonucleotides, as a control, axonal preference (100%, 122 boundaries were evaluated, n=12) for the ALCAM-containing lanes is fully maintained (Fig. 8C). Taken together, the experiments show that an undisturbed ALCAM translation (i.e. a physiological protein level in the growth cone plasma membrane) is necessary for the restriction of RGC axon growth along ALCAM-presenting pathways.

Because the ALCAM protein is present on projecting axons and absent from their somata during nervous system development (Pourquie et al., 1992; Pollerberg and Mack, 1994), we investigated whether ALCAM might be translated locally in axons and found six independent lines of evidence that are, furthermore, supported by other studies: (1) We found ALCAM mRNA in growth cones (in vitro and in vivo). The mRNA of another IgSF-CAM (NCAM) has been found in axons and growth cones (in vitro) but the function was not further elucidated (Brittis et al., 2002). (2) Two different reporter proteins are translated in growth cones when under the control of the ALCAM 3′-UTR. These reporters (dGFP, pTimer) have also been used to demonstrate growth cone translation in other studies (Brittis et al., 2002; Willis et al., 2007; Besse and Ephrussi, 2008). The remaining reporter signal in growth cones with ALCAM 3′-UTR mutations, antisense treatment or inhibitor application, as well as the remaining ALCAM protein levels in the growth cone plasma membrane after antisense treatment, also indicate the presence of a somatic ALCAM production. Both, the reporter data and the analysis of cut growth cones show, however, that the somatically produced ALCAM does not substantially contribute to the ALCAM protein levels in the growth cone plasma membrane. (3) The level of ALCAM protein is maintained in cut growth cones (if translation is not inhibited), in contrast to the level of another IgSF-CAM, CAML1, the levels of which decreases (independent of translation inhibition) in cut growth cones. This is in accordance with the finding that presence of CAML1 in growth cones requires its synthesis and sorting in the somatic compartment (Kamiguchi and Lemmon, 1998; Wisco et al., 2003; Yap et al., 2008). ALCAM protein lacks sequences for axonal sorting and targeting such as found in CAML1 (Kamiguchi and Lemmon, 1998; Yap et al., 2008). (4) Mutations in CPEs in the 3′-UTR of ALCAM strongly reduce its growth cone translation; it has been shown that CPEs are capable of transporting mRNA in neuronal fibers (Huang et al., 2003). Moreover, there are no sequences in the flanking regions of ALCAM mRNA pointing to alternative transport systems, together indicating that the CPEs might also support the axonal transport of ALCAM mRNA – in addition to regulating its translation. CPEs have indeed been shown to possess dual roles in transport and translation of mRNAs (Huang et al., 2003), which is controlled by CPE-binding proteins (CPEBs) that are also present in the retina (McKee et al., 2005). Binding of CPEBs to the CPEs allows for the transport of the mRNA, while at the same time repressing its translation (Richter, 2007). Triggered by an external stimulus, CPEBs are phosphorylated, which stimulates growth cone translation; conceivably, the CPEB phosphorylation could be a limiting factor in ALCAM translation. CPEB phosphorylation is crucial for appropriate growth cone reactions to the environment (Brittis et al., 2002; Kundel et al., 2009) and thereby might also be important for the impact of ALCAM on axonal navigation (see below). (5) Inhibition of ERK or TOR abolishes ALCAM-dependent translation; both kinases have been shown to be a pre-requisite for growth cone translation and are activated in growth cones by extracellular stimuli (Zhou and Snider, 2006; Jung and Holt, 2011). Thus the onset, level and growth cone subregion of ALCAM synthesis might be regulated by environmental cues that stimulate these two kinases. It is noteworthy in this context that ERK inhibition reduces not only ALCAM's the growth cone translation of ALCAM but also its endocytosis and degradation (Thelen et al., 2008); therefore, the reduced axonal elongation and navigation, observed upon ERK inhibition, is rather due to a reduced re-sensitization of ALCAM (see below) and not to changed ALCAM levels in the cell membrane. (6) The rapid re-appearance of ALCAM in the growth cone plasma membrane (10 minutes after triggering of endocytosis) is prevented by inhibition of its translation. Replacement of growth cone ALCAM by somatically produced ALCAM can be basically excluded because somatic synthesis and folding of large plasma membrane proteins requires at least 10 minutes (Lodish et al., 1983; Yewdell, 2001). In addition, targeting, sorting and filtering at the axon initial segment are impediments for a (swift) appearance of membrane proteins in the axon (Song et al., 2009). The endocytosis of ALCAM in the growth cone results in its degradation and not in recycling (Thelen et al., 2008), also indicating that there is a need for a local synthesis of this CAM. In vivo, ALCAM endocytosis could be triggered by contact of RGC growth cones with ALCAM-presenting RGC axons. This homophilic trans-interaction can only take place in the optic fiber layer and might thus – owing to the rapid activation of ALCAM translation – keep the RGC growth cones in this layer. Translation of ALCAM in growth cones also conceivably makes a rapid stop of this local synthesis possible, because no signal has to be sent to the soma to stop ALCAM production and no axonal pool of (just transported) ALCAM has to be dealt with. RGC growth cones on their way to the mesencephalon might immediately cease ALCAM translation upon contact with the chiasm, as growing RGC axons cease to display ALCAM (in contrast to CAML1) after passing the chiasm (Pollerberg and Mack, 1994; Buhusi et al., 2009), which might be of crucial importance for their path and target finding capability.

The complete omission of the axon elongation enhancement induced by substrate ALCAM, which was achieved by the block of its growth cone translation, reveals that the interaction partner for ALCAM in the growth cone plasma membrane is ALCAM, i.e. a homophilic trans-interaction. In vivo, this enhancement might selectively support those RGC axons that have found the optic fiber layer (the only ALCAM-presenting retina layer) by accelerating their advance and preventing misrouting. A critical density of trans-interactions with substrate ALCAM is necessary for axon growth, as shown previously using substrates with controlled ALCAM nano-spacing (Jaehrling et al., 2009). The present study demonstrates that ALCAM levels that are too high or low in the plasma membrane (achieved by overexpression or knockdown) abolish the substrate ALCAM-induced axon growth enhancement, indicating that a physiological ALCAM density in the axon membrane is at a fine-tuned optimum. Overexpression of ALCAM cannot be achieved in neurons that endogenously express ALCAM, indicating that the 3′-UTR-mediated translation repression only occurs in such neurons. This again points to the importance of the prevention of too high ALCAM levels in neurons as this might either cause too strong adhesion, trigger unphysiological levels of intracellular signaling or disable growth cone re-sensitization (Piper et al., 2005).

The ALCAM protein level in the plasma membrane is not only downregulated by translation repression in the growth cone but also by local endocytosis leading to degradation of ALCAM (Thelen et al., 2008). The CAML1, by contrast, is (upon homo- and heterophilic interactions) internalized and recycled in growth cones (Kamiguchi and Yoshihara, 2001; Castellani et al., 2004). ALCAM-elicited ERK activation decreases fast – including under blockage of endocytosis – indicating that there is a progressive loss in the ability of ALCAM to trigger intracellular signaling (Thelen et al., 2008). These ‘de-sensitized’ ALCAM proteins conceivably have to be removed from the plasma membrane and replaced by newly synthesized, signaling-competent ALCAM (‘re-sensitization’) to ensure a rapid regain of responsiveness (to ALCAM trans-interaction partners) of the growth cone, as has been proposed for other growth cone proteins (Piper et al., 2005). Our finding that, upon activation, ALCAM is both translated and endocytosed in the growth cone indicates that a rapid and well-balanced replacement of ALCAM is necessary in order to keep the growth cone functioning. Endocytosis and translation of ALCAM, as well as ERK activation, are each by itself crucial for the ability of the growth cone to prefer ALCAM-containing lanes. ERK appears to be a key player in the regulation of ALCAM plasma membrane levels because this kinase, upon ALCAM activation, stimulates endocytosis as well as translation of ALCAM, with both processes acting in concert to ensure re-sensitization of ALCAM. The concomitant regulation of both processes by ERK might ensure the coordinated, swift control of ALCAM density in the plasma membrane and prevent fluctuations leading to fatal routing errors of the rapidly advancing growth cone, as have been observed upon transient, as well as permanent, abolishment of ALCAM in the developing retina (Avci et al., 2004; Weiner et al., 2004).

Here, we show that the preference of RGC axons for ALCAM-containing substrate lanes requires not only the presence of ALCAM protein but its continuous synthesis. These experiments, moreover, show that this preference is mediated by homophilic ALCAM trans-interactions and not by heterophilic ones (DeBernardo and Chang, 1996; Buhusi et al., 2009). ALCAM trans-interactions, which are only possible in the growth cone subregion situated on the ALCAM-presenting lane, might result in clustering, activation, ubiquitylation, endocytosis and degradation of ALCAM (Thelen et al., 2008), as well as its local translation and membrane insertion. This might lead to new cycles of ALCAM activation and intracellular signaling in this growth cone subregion, enhancing the local cytoskeleton-based stabilization that causes protrusions at this site; ERK has indeed been shown to regulate cytoskeletal dynamics (Wortzel and Seger, 2011). Hence, local ALCAM activation might – through the elicited signaling affecting downstream cytoskeletal components – ensure that there are adequate growth cone responses, resulting in appropriate steering maneuvers and ultimately proper axon navigation.

The cDNA encoding chick ALCAM was a gift from Hideaki Tanaka (Kumamoto University, Honjo, Japan), destabilized GFP with the γ-actin 3′-UTR (pcDNA3-dGFP-γactin) was a gift from Jefferey Twiss (Alfred I. duPont Hospital for Children, Wilmington, Delaware). The coding sequence of ALCAM was subcloned into the expression plasmid pSecTag (Invitrogen). CPE mutations in the 3′-UTR of ALCAM were prepared with the Altered Sites II site-directed mutagenesis kit (Promega) according to the manufacturer's specifications by using 5′-AGCCTTACAGTGTATCCGAATAGCTACGTAAAA-3′ (CPE1) and 5′-CCCAAAGGGTGTCATCCGAACAATTGAAATCAA-3′ (CPE2) oligonucleotides (Sigma); mutations were confirmed by DNA sequencing. Wild-type and mutant 3′-UTRs were subcloned into the NotI/EcoRI site of pTimer (Clontech). The γ-actin 3′-UTR of pcDNA3-dGFP-γactin was cut using NotI and XhoI, the wild-type 3′-UTR of ALCAM was amplified by PCR and inserted into the NotI/XhoI site of pcDNA3-dGFP-γactin devoid of the γ-actin 3′-UTR. For a short hairpin (sh)RNA-mediated knockdown of ALCAM, the DNA sequence 5′-GATCCGTACCAGATGGTCTGATGTTTCAAGAGAACATCAGACCATCTGGTACTTTTTTGGAAA-3′ was inserted into pSilencer (Ambion). CPE inhibition oligonucleotides were as follows: CPE-1 sense: 5′-TAGCTATTTTTATACACTG-3′; CPE-1 antisense: 5′-CAGTGTATAAAAATAGCTA-3′; CPE-2 sense: 5′-CAATTGTTTTTATGACACC-3′; CPE-2 antisense: 5′-GGTGTCATAAAAACAATTG-3′.

E6 retinal explants (Avci et al., 2004) were placed onto coverslips coated with PLL (40 μg/ml) and laminin (5 μg/ml) and incubated for 2 days at 37°C under 5% CO₂. For RT-PCR, RNA was extracted from the growth cone and distal axon fraction, and the explant fraction directly on the coverslip in 15 μl lysis buffer, transferred into a microcentrifuge tube, and reverse transcribed by using the First Strand cDNA Synthesis Kit (MBI Fermentas) using random primers. Qualitative PCR used Taq polymerase with hot start at 92°C (2 minutes), 94°C (2 minutes), then 50 cycles at 95°C (30 seconds), 56°C (30 seconds), 72°C (1 minute), and finally 72°C (10 minutes). Primers used were ALCAM-forward (5′-CCTGTGTCGTGCACAATTTC-3′, nucleotides 1065–1085; ALCAM-reverse 5′-AACTGTCCATTGCACTGCTG-3′, 1557–1574), β-actin-forward (5′-GACTGTTACCAACACCCACACC-3′, 1201–1222), β-actin-reverse (5′-CTTCACAGAGGCGAGTAACTTC-3′, 1471–1494), histone H3-forward (5′-AAGAAGCCTCACCGCTACAG-3′, 109–129), and histone-reverse (5′-GCGTATCTGGTGGGTCTGTT-3′, 295–315).

Sense and antisense riboprobes (543 bp) were in vitro transcribed using the DIG RNA Labeling Kit (SP6/T7) (Roche) with digoxigenin-conjugated UTP. Cell cultures (1 day in vitro) of E6 chick retinae were fixed for 60 minutes in 4% paraformaldehyde (PFA) in PBS. Cells were permeabilized in 0.2% Tween20 in PBS (10 minutes), treated with 0.3 U/ml Proteinase K and 5 mM EDTA (3 minutes), washed with PBS and fixed again (5 minutes). Cultures were then equilibrated 3× with 0.1 M Triethanolamin buffer (pH 8.0) and incubated with fresh acetylation buffer (0.25% acetic anhydride in Triethanolamin buffer, 10 minutes). Next, cells were equilibrated with hybridization buffer (50% formamide, 10 mM Tris-HCl, 250 μg/ml yeast tRNA, 1× Denhardt's solution, 600 mM NaCl, 0.25% SDS, 1 mM EDTA, 10% dextran sulfate) for 1–3 hours and incubated with 500 ng riboprobes in 50 μl hybridization buffer at 60°C overnight. The coverslips were washed twice with 2× SSC at 60°C (20 minutes) and incubated at 37°C with 12.5 μg/ml RNase A in 2× SSC (30 minutes). Coverslips were then washed twice with 2× SSC at 37°C, 3× with PBS, 1× with 0.2% Tween20 in PBS and finally 2× with PBS. Cultures were blocked with 0.5% Blocking Reagent (Roche) in PBS (30 minutes) and hybridization was detected with anti-digoxigenin antibody (Roche, 1:200 in Blocking Reagent, 1 hour) followed by a goat-anti-mouse–Alexa-Fluor-488 antibody (1:250). All micrographs were taken using the same acquisition conditions with an Axiovert 200M (Zeiss) equipped with a 63× objective (NA 1.2) and a digital camera AxioCam (Zeiss).

12 μm sections of the E6 retina were dried at 50°C, fixed with 4% PFA/PBS (20 minutes) and washed twice with PBS (5 minutes). Sections were treated with 5 μg/ml Proteinase K (10 minutes), washed 1× with PBS, fixed with 4% PFA in PBS (15 minutes) and washed 1× with distilled water. Tissue sections were then incubated with fresh acetylation buffer (0.25% acetic anhydride in Triethanolamin buffer, 10 minutes) at room temperature and washed once with PBS. Next, sections were equilibrated with hybridization buffer (70°C, 4 hours) and incubated with 500 ng riboprobes in 100 μl hybridization buffer (70°C, overnight). The slides were washed once with 2× SSC (70°C, 20 minutes) and once with 2× SSC (room temperature, 5 minutes) before incubation with 10 ng/ml RNase A and 0.2 U RNase T1 (Sigma) in 2× SSC (37°C, 30 minutes) followed by washing twice with 2× SSC. Slides were then washed 2× with 0.2× SSC (70°C, 30 minutes), 1× in 0.2× SSC (room temperature, 2 minutes) and twice with 0.2% Tween20 in PBS (20 minutes) and finally 2× with PBS. Cultures were blocked with blocking buffer (10% horse serum in PBS, 1 hour) and hybridization was detected with alkaline-phosphatase-conjugated anti-digoxigenin antibody (1:200 in blocking buffer, 1 hour) followed alkaline phosphatase reaction in NBT/BCIP solution (Roche). All animal experiments were performed according to approved guidelines.

Heads of chick embryos were fixed, sectioned, and stained as described previously (Pollerberg et al., 1985; Zelina et al., 2005). Images were taken using an inverted fluorescence microscope Axiovert 200M (Zeiss) equipped with a digital AxioCam camera (Zeiss). Pictures were processed with the Axiovision 4.2 software and Image J (NIH) and annotated with Adobe Photoshop CS2.

E6 retinal explants were grown on laminin-coated coverslips (20 μg/μl). 24 hours after the start of the explant culture, distal axons were separated from the proximal axons by cutting with a fine tungsten wire and cultured for additional 6 hours before fixation with 4% PFA. Growth cones of dissected axons survived for more than 8 hours; 12 hours after dissection, however, no growth cones could be observed. For inhibition of protein translation, cycloheximide (Calbiochem, 35 μM) was added to the culture medium directly after axon dissection.

Cell cultures of E6 chick retinae were prepared as described previously (Pollerberg and Mack, 1994). For transfection, 4×10⁶ retinal cells in 100 μl DMEM/F12 were mixed with 7 μg DNA and electroporated using the Nucleofector device (Amaxa Biosystems) using the kit VPG 1002 and program G-13. 1×10⁵ cells were plated per coverslip coated with poly-L-lysine (40 μg/ml; Sigma) and laminin (20 μg/ml, Invitrogen) and fixed after 36 hours. Axon length was measured using the NeuronJ plugin of ImageJ (NIH). Axon preference assays with retinal explant strips of E6 retina were prepared as described previously (Avci et al., 2004) except that recombinant, soluble ALCAM (3 μg/ml) was employed (Thelen et al., 2007). 24 hours after the start of the explant culture, U0126 (10 μM, an inhibitor of the ERK-activating kinase MEK, Calbiochem) or oligonucleotides [5 μM (Leung et al., 2006; Yao et al., 2006)] were added, followed by an additional 24-hour incubation before fixation with 4% PFA. A substrate lane was counted as exerting axonal preference when it contained axons or axon bundles and at the same time contained no axons in the neighboring lanes (or only a very few axons). A retinal explant strip was considered to be showing preference if its RGC axons respected the borders of at least 50% of the lanes of a given substrate. Images were taken and assembled with a Nikon Eclipse 90i equipped with a Nikon DS-1QM camera (Nikon Imaging Center, University of Heidelberg) using the scan large image function.

Clustering of plasma membrane ALCAM and quantification were performed as described previously (Thelen et al., 2007). To visualize all ALCAM proteins (clustered and non-clustered) present on the cell surface, cells were labeled with the same polyclonal anti-ALCAM antibody as used for clustering. To exclude the impact of soma-synthesized ALCAM, only growth cones of axons longer than 100 μm were evaluated.

Fluorescent images were captured using constant exposure times and avoiding pixel saturation. For quantification of fluorescence intensity (FI) of surface ALCAM, CAML1, mRNA detected by FISH and dGFP, the growth cone outlines were traced in the phase contrast images using ImageJ, superimposed on the fluorescent images and the FI within the growth cone was determined. To quantify young and aged protein pTIMER, green and red FIs of transfected retinal ganglion cells (having formed axons with a minimal length of 150 μm) were measured in the soma, in the proximal axon (20 μm behind the transition of soma to axon), in the medial axon (halfway between soma and growth cone), in the distal axon (20 μm before the growth cone) and in the growth cone (starting when reaching a diameter of twice as the axon). The background FIs were measured in an adjacent area clear of cellular material and subtracted, yielding the background-corrected intensity. To analyze the FI of the young (green) reporter signals in relation to total reporter signal, the signals of aged (red) and young reporter were added together, set to 100% (normalizing the different FI levels of the individual RGCs) and the contribution of the young reporter was calculated.

pMiw-ALCAM and pEGFP-N1 (2 μg/μl each) were mixed with Fast Green (1%) to facilitate visualization of the injected DNA solution. DNA solution (~30 nl) was injected into the optic vesicles of Hamburger–Hamilton stage 10 embryos. After the injection, the anode was placed beside the right optic vesicle and the cathode beside the left optic vesicle. An electric pulse was applied for 50 mseconds, three times at 20 V. Electroporated embryos were incubated at 38.5°C until they reached embryonic day 4.5.

We thank S. Bergmann, C. Brandel, and M. Zieher-Lorenz for excellent technical assistance. We are grateful to H. Tanaka (Kumamoto University, Honjo, Japan) for pMIW-ALCAM and J. Twiss (Alfred I. duPont Hospital for Children, Wilmington, Delaware, USA) for the destabilized GFP construct. We thank H. Avci (CNRS, Paris, France) for stimulating discussions and are grateful to U. Engel and C. Ackermann (Nikon Imaging Center at the University of Heidelberg) for their advice.