Vertebrate somitogenesis is regulated by a segmentation clock. Clock-linked genes exhibit cyclic expression, with a periodicity matching the rate of somite production. In mice, lunatic fringe (Lfng) expression oscillates, and LFNG protein contributes to periodic repression of Notch signaling. We hypothesized that rapid LFNG turnover could be regulated by protein processing and secretion. Here, we describe a novel Lfng allele (LfngRLFNG), replacing the N-terminal sequences of LFNG, which allow for protein processing and secretion, with the N-terminus of radical fringe (a Golgi-resident protein). This allele is predicted to prevent protein secretion without altering the activity of LFNG, thus increasing the intracellular half-life of the protein. This allele causes dominant skeletal and somite abnormalities that are distinct from those seen in Lfng loss-of-function embryos. Expression of clock-linked genes is perturbed and mature Hes7 transcripts are stabilized in the presomitic mesoderm of mutant mice, suggesting that both transcriptional and post-transcriptional regulation of clock components are perturbed by RLFNG expression. Contrasting phenotypes in the segmentation clock and somite patterning of mutant mice suggest that LFNG protein may have context-dependent effects on Notch activity.
INTRODUCTION
During somitogenesis, somites bud from the presomitic mesoderm (PSM) and give rise to the axial skeleton and other adult structures. In the PSM, several genes exhibit oscillatory expression, with a cyclic period that matches the rate of somite formation, linking them to the segmentation clock. Many of these genes are members of the Notch pathway, although genes in the FGF and WNT pathways also exhibit cyclic expression (reviewed by Kageyama et al., 2012). Notch activity levels in the PSM also oscillate (Morimoto et al., 2005; Shifley et al., 2008), and in mouse and chicken this requires periodic repression of Notch signaling by lunatic fringe (Lfng), which functions in a delayed negative-feedback loop with the transcriptional repressor Hes7 to regulate cyclic Notch activation in the clock (Bessho et al., 2003; Kageyama et al., 2012).
In mouse and chick embryos, cyclic Lfng transcription and oscillation of LFNG protein levels have been observed, linking cyclic LFNG activity to the clock in the posterior PSM (Cole et al., 2002; Dale et al., 2003; Morales et al., 2002). The short period of these oscillations predicts that post-translational mechanisms must exist to enforce a short LFNG protein half-life, allowing rapid protein turnover during the ‘off' phases of the clock cycle. Supporting this idea, it has been shown that either loss of Lfng activity (Evrard et al., 1998; Zhang and Gridley, 1998) or sustained, non-oscillatory Lfng expression in the PSM (Serth et al., 2003) perturbs embryonic segmentation and clock function. Lfng is also expressed in the anterior PSM, in the rostral compartment of patterning presomites (Cole et al., 2002), although most data suggest that its most important roles in segmentation are in the clock (Oginuma et al., 2010; Williams et al., 2014).
Lfng encodes a glycosyltransferase (O-fucosylpeptide 3-beta-N-acetylglucosaminyltransferase) that modifies Notch and its ligands in the Golgi, modulating ligand-receptor interactions (Moloney et al., 2000; Panin et al., 2002; Serth et al., 2015). Interestingly, the effects of LFNG protein on Notch signaling are context dependent. In most systems, LFNG acts in the signal-receiving cell, where it potentiates the activation of Notch receptors by Delta-like ligands while reducing signaling by Jagged ligands (Hicks et al., 2000; Kato et al., 2010; Yang et al., 2005). However, in the clock it has been suggested that LFNG protein may act in the signal-sending cell to inhibit activation of NOTCH1 by DLL1, playing a crucial role in the synchronization of clock oscillations between neighboring cells (Okubo et al., 2012).
In most cases, the described functions of LFNG in the Notch pathway are cell-autonomous (Hicks et al., 2000; Moloney et al., 2000; Munro and Freeman, 2000; Okubo et al., 2012; Panin et al., 1997). However, both Drosophila Fringe and LFNG are cleaved by furin-like proteases and secreted into the extracellular space (Johnston et al., 1997; Shifley and Cole, 2008), where they may be inactive. Accelerating the secretion of Drosophila Fringe creates a hypomorphic allele, supporting the idea that protein processing and secretion limit its cell-autonomous activity (Munro and Freeman, 2000). We hypothesized that processing and secretion of LFNG terminates its intracellular activity, facilitating the rapid turnover required for its function in the segmentation clock, and rapidly clearing the protein from the posterior compartments of presomites in the anterior PSM. This would represent a novel post-translational mechanism for the spatial and temporal regulation of Notch signaling, allowing rapid and reversible modulation of Notch activity.
To examine this hypothesis, we previously analyzed a chimeric fusion of the active domain of LFNG to the N-terminal sequences of radical fringe (RFNG), which is a Golgi-resident fringe protein. When this mutant protein was expressed in tissue culture cells under the control of an exogenous promoter, we found that the protein was targeted to the Golgi and maintained the enzymatic activity of LFNG (Shifley and Cole, 2008). However, unlike wild-type LFNG protein, the chimeric protein was not secreted from tissue culture cells even when strongly overexpressed. Furthermore, the intracellular half-life of the chimeric protein was twice as long as that of wild-type LFNG (Shifley and Cole, 2008). Here, we generate mice expressing this Golgi-resident RFNG/LFNG chimeric variant (RLFNG) from the endogenous Lfng locus by replacing the N-terminal sequences of LFNG, which contain the protein processing sites, with the N-terminus of RFNG. Based on our previous findings in tissue culture, this alteration is predicted to prevent protein processing and secretion, while maintaining the enzymatic activity and specificity of LFNG.
We observe severe segmentation defects in heterozygous mutant mice that express RLFNG protein, and these are distinct from those observed in animals lacking LFNG activity. These outcomes are consistent with the hypothesis that altering protein processing/secretion creates a hypermorphic RLFNG protein by preventing its rapid turnover in the Golgi. Clock-related phenotypes suggest that Notch activity levels are decreased in the posterior PSM of mutant embryos. By contrast, somite phenotypes in mutant mice are suggestive of expanded Notch signaling during rostral/caudal patterning of somites. These findings support the idea that LFNG protein may have context-dependent effects on Notch signaling in different tissue compartments. Finally, we observe that this mutation affects the transcript stability of another clock component, Hes7, suggesting that there may be links between normal clock function and the post-transcriptional regulation of crucial clock components.
RESULTS
Expression of RLFNG protein perturbs segmentation
We previously described a chimeric RFNG/LFNG protein (RLFNG) that is localized to the Golgi, tethered in the cell, maintains the enzymatic activity of endogenous LFNG, but has a longer intracellular half-life than the wild-type LFNG protein (Shifley and Cole, 2008). We expressed RLFNG in vivo by replacing exon 1 of the endogenous Lfng locus with a new exon containing sequences encoding the N-terminus and type II transmembrane domain of RFNG fused to LFNG at the first conserved amino acid of the proteins (Fig. 1A, Fig. S1). Based on our results in tissue culture, we predict that the resulting allele, LfngRLFNG, encodes a protein with the enzymatic activity of LFNG but which is not secreted. This allele does not alter the promoter or enhancer regions, transcription start site, UTRs, or splice sites of Lfng, and thus the mutation is not predicted to have direct effects on Lfng gene expression. If LFNG secretion from wild-type cells acts to inactivate LFNG protein activity in the clock, and the mouse allele prevents this secretion as was seen in tissue culture, we predict that this allele will inhibit inactivation of the LFNG intracellular activity. This would result in a functionally hypermorphic RLFNG protein that is retained and active in the Golgi longer than the wild-type protein. Owing to its longer intracellular half-life, RLFNG would not be cleared from cells during the brief ‘off' phase of the clock. Persistent LFNG activity would then be expected to perturb the clock feedback loop, actively interfering with normal clock function and somitogenesis. We predicted that this allele would be associated with dominant phenotypes, as wild-type LFNG protein cannot compensate for the inappropriate perdurance of the RLFNG protein.
Skeletal abnormalities in mice expressing RLFNG protein. (A) The endogenous Lfng locus was targeted to replace the Lfng pre/pro region and 28 amino acids of the mature region (gray boxes and rectangle) with the signal sequence and 21 amino acids of the mature domain (striped box and rectangle) of RFNG (full targeting plan provided in Fig. S1). A Southern blot of targeted colonies with a 3′ flanking probe (arrow indicates endogenous band, arrowhead indicates targeted band) is shown. (B) Heterozygous mice (right) are viable but have a shortened body axis and truncated tail. (C) Skeletons of 17.5 dpc wild-type (a-c; n=10), Lfng null (d-f; n=9) and LfngRLFNG/+(g-i; n=12) embryos stained with Alizarin Red and Alcian Blue. LfngRLFNG/+ mice have more severe rib fusions (arrow in g) and more severe neural arch fusions (asterisk in g and i) than those observed in Lfng null mice. In the sacral region, Lfng null mice have one to four relatively normal vertebrae (arrowheads in f), in contrast to LfngRLFNG/+ animals, which have disorganized sacral regions, ectopic ossifications (arrow in i) and arch fusions (asterisk in i).
Skeletal abnormalities in mice expressing RLFNG protein. (A) The endogenous Lfng locus was targeted to replace the Lfng pre/pro region and 28 amino acids of the mature region (gray boxes and rectangle) with the signal sequence and 21 amino acids of the mature domain (striped box and rectangle) of RFNG (full targeting plan provided in Fig. S1). A Southern blot of targeted colonies with a 3′ flanking probe (arrow indicates endogenous band, arrowhead indicates targeted band) is shown. (B) Heterozygous mice (right) are viable but have a shortened body axis and truncated tail. (C) Skeletons of 17.5 dpc wild-type (a-c; n=10), Lfng null (d-f; n=9) and LfngRLFNG/+(g-i; n=12) embryos stained with Alizarin Red and Alcian Blue. LfngRLFNG/+ mice have more severe rib fusions (arrow in g) and more severe neural arch fusions (asterisk in g and i) than those observed in Lfng null mice. In the sacral region, Lfng null mice have one to four relatively normal vertebrae (arrowheads in f), in contrast to LfngRLFNG/+ animals, which have disorganized sacral regions, ectopic ossifications (arrow in i) and arch fusions (asterisk in i).
Indeed, heterozygous (LfngRLFNG/+) mice are viable and fertile, but exhibit skeletal abnormalities, with a shortened body axis and truncated tail, outwardly similar to Lfng null animals (Fig. 1B). However, the phenotypes observed in LfngRLFNG/+ mice are distinct from those described in Lfng null mutants, as expected for a gain-of-function mutation. In the anterior skeleton, LfngRLFNG/+ animals exhibit increased fusions of dorsal ribs and neural arches compared with Lfng null animals (Fig. 1C, compare d,e with g,h). The sacral region is severely affected in LfngRLFNG/+ animals, with neural arch fusions and disorganized and ectopic ossification centers. This is in contrast to Lfng null embryos, which form one to four sacral condensations with relatively normal neural arch morphology (Fig. 1C, compare f with i). Thus, the expression of a chimeric protein that is expected to prevent LFNG secretion creates a dominant gain-of-function allele that perturbs skeletal development more severely than the complete loss of LFNG function.
Endogenous Lfng transcription is reduced in LfngRLFNG/+ embryos
In the clock, LFNG functions in a feedback loop with NOTCH1 and Hes7, and thus changes in LFNG protein turnover are predicted to act through this loop to affect the expression of Lfng in the posterior PSM. However, Lfng expression in the anterior PSM is driven by distinct regulatory elements that may be independent of these feedback loops (Cole et al., 2002). To examine the effects of the RLFNG protein on Lfng expression, we examined LfngRLFNG/+ embryos at 10.5 days post coitum (dpc) and observed expression of Lfng mRNA in a weak stripe in the anterior PSM, with no expression of Lfng mRNA in the posterior PSM (Fig. 2Aa-d). Lfng expression in the posterior PSM was examined more closely in 8.5 dpc mutant embryos, where background staining is less noticeable and thus lower levels of expression can be reliably detected. After a 6-h incubation in detection solution, cyclic Lfng expression was clearly visible in wild-type embryos, but only a faint band in the anterior PSM was seen in LfngRLFNG/+ embryos (Fig. 2Ae-g). However, after a 48-h incubation in detection solution, Lfng expression was observed throughout the posterior PSM of LfngRLFNG/+ embryos (Fig. 2Ah). This expression is not overtly oscillatory, although the low expression levels make it difficult to exclude the possibility of dynamic transcription. However, dynamic expression in the caudal PSM of wild-type embryos was evident after this prolonged detection (Fig. S2A), suggesting that the lack of overt cycling in LfngRLFNG/+ embryos is unlikely to be an artifact of the long detection. qRT-PCR on RNA from individual caudal PSM confirms that LfngRLFNG/+ embryos express relatively constant low levels of Lfng transcript comparable to the lowest levels of Lfng expression observed in wild-type embryos (Fig. S2B). Standard PCR and qRT-PCR using allele-specific primers confirms that both the mutant and wild-type alleles are expressed to similar levels in mutant embryos (Fig. S2C).
Lfng expression and Notch activation are perturbed in LfngRLFNG/+ embryos. (A) Lfng expression in 10.5 dpc wild-type embryos oscillates (a-c; n=4/13 phase 1, 5/13 phase 2, 4/13 phase 3), but is reduced in the posterior PSM of LfngRLFNG/+ embryos (d; n=10). At 8.5 dpc cyclic Lfng expression is observed in wild-type embryos after 6 h of detection (e,f; n=4/10 phase 1, 6/10 phase 2), but Lfng expression in the posterior PSM of LfngRLFNG/+ embryos is not observed (g; n=7). After 48 h of detection, low levels of Lfng are seen in the posterior PSM of LfngRLFNG/+ embryos (h; n=4). (B) Immunohistochemistry for NICD demonstrates reduced Notch activity in the posterior PSM of LfngRLFNG/+ embryos (d; n=9) compared with wild-type embryos (a-c; n=3/12 phase 1, 6/12 phase 2, 3/12 phase 3). Long exposures at 8.5 dpc reveal low levels of NICD in the caudal PSM of mutant embryos (g; n=6), but cyclic NICD expression in wild-type embryos (e,f; n=4). Longer exposures of 9.0 dpc embryos reveal that NICD is confined to the caudal compartments of the presomites (arrows) and mature somites (asterisks) in wild-type embryos (h; n=4), whereas in LfngRLFNG/+ embryos (i; n=4) a broad band of NICD is seen in the anterior PSM (arrow), with low levels throughout the mature somites (line). In h and i the images focus on the anterior PSM and the most recently formed somites. (C) Nrarp expression is reduced in the PSM of LfngRLFNG/+ embryos (d; n=9) compared with wild-type embryos (a-c; n=4/11 phase 1, 3/11 phase 2, 4/11 phase 3). (D) There are no dramatic differences in the expression of Notch1 (a,b) and Dll1 (c,d) between wild-type (a, n=3; c, n=5) and LfngRLFNG/+ (b, n=3; d, n=4) embryos. For oscillatory genes, the distributions of expression patterns were significantly different between wild-type and mutant embryos by Fisher's exact analysis (P<0.01).
Lfng expression and Notch activation are perturbed in LfngRLFNG/+ embryos. (A) Lfng expression in 10.5 dpc wild-type embryos oscillates (a-c; n=4/13 phase 1, 5/13 phase 2, 4/13 phase 3), but is reduced in the posterior PSM of LfngRLFNG/+ embryos (d; n=10). At 8.5 dpc cyclic Lfng expression is observed in wild-type embryos after 6 h of detection (e,f; n=4/10 phase 1, 6/10 phase 2), but Lfng expression in the posterior PSM of LfngRLFNG/+ embryos is not observed (g; n=7). After 48 h of detection, low levels of Lfng are seen in the posterior PSM of LfngRLFNG/+ embryos (h; n=4). (B) Immunohistochemistry for NICD demonstrates reduced Notch activity in the posterior PSM of LfngRLFNG/+ embryos (d; n=9) compared with wild-type embryos (a-c; n=3/12 phase 1, 6/12 phase 2, 3/12 phase 3). Long exposures at 8.5 dpc reveal low levels of NICD in the caudal PSM of mutant embryos (g; n=6), but cyclic NICD expression in wild-type embryos (e,f; n=4). Longer exposures of 9.0 dpc embryos reveal that NICD is confined to the caudal compartments of the presomites (arrows) and mature somites (asterisks) in wild-type embryos (h; n=4), whereas in LfngRLFNG/+ embryos (i; n=4) a broad band of NICD is seen in the anterior PSM (arrow), with low levels throughout the mature somites (line). In h and i the images focus on the anterior PSM and the most recently formed somites. (C) Nrarp expression is reduced in the PSM of LfngRLFNG/+ embryos (d; n=9) compared with wild-type embryos (a-c; n=4/11 phase 1, 3/11 phase 2, 4/11 phase 3). (D) There are no dramatic differences in the expression of Notch1 (a,b) and Dll1 (c,d) between wild-type (a, n=3; c, n=5) and LfngRLFNG/+ (b, n=3; d, n=4) embryos. For oscillatory genes, the distributions of expression patterns were significantly different between wild-type and mutant embryos by Fisher's exact analysis (P<0.01).
The LfngRLFNG allele contains endogenous transcriptional regulation sequences that are, in principle, sufficient to drive expression of the RLFNG protein in a wild-type pattern. However, our results suggest that the RLFNG protein expressed from the mutant allele perturbs the feedback loop that controls Lfng transcription, leading to altered clock function and secondary effects on Lfng transcription in the posterior PSM. Importantly, this reduction in Lfng transcripts is unlikely to be the direct cause of the phenotypes observed in our mutant mice, as phenotypes associated with the LfngRLFNG allele are dominant, and do not recapitulate the Lfng loss-of-function phenotypes. Instead, our results support the hypothesis that this low level of transcript produces sufficient stabilized RLFNG protein to continuously perturb clock function.
RLFNG expression perturbs cyclic activity in the Notch pathway
In the segmentation clock, LFNG protein has been proposed to inhibit Notch activation (Okubo et al., 2012). However, in the anterior PSM, Lfng expression is confined to the future rostral somite compartment, whereas Notch is activated in the future caudal compartment (Morimoto et al., 2005). To examine the effect of the LfngRLFNG allele on Notch signaling in the PSM, we examined NOTCH1 activation by whole-mount immunohistochemistry with an antibody specific to cleaved NOTCH1 [Notch intracellular domain (NICD)]. In wild-type embryos, NICD levels oscillate in the posterior PSM, and are stable in a stripe in the anterior PSM, as previously described (Morimoto et al., 2005; Shifley et al., 2008). At identical detection times, NICD is seen only in a stripe in the anterior PSM of LfngRLFNG/+ embryos (Fig. 2Ba-d). After longer detection we observe low levels of NICD in the posterior PSM of LfngRLFNG/+ embryos (Fig. 2Bg), which are not overtly oscillatory, as well as a broad band of NICD in the anterior PSM and low levels of ubiquitous NICD in the somitic paraxial mesoderm. This is in contrast to the defined stripes of NICD observed in the caudal compartments of presomites in the anterior PSM and epithelial somites of wild-type embryos (Fig. 2Bh,i).
This pattern of Notch activity is confirmed by the expression of Nrarp, a direct Notch target in the PSM (Dequeant et al., 2006; Wright et al., 2009), which is severely reduced in the posterior PSM of all LfngRLFNG/+ embryos examined, with expression restricted to an anterior stripe (Fig. 2Ca-d). These phenotypes are different from those observed in Lfng null embryos, which exhibit stable NICD levels throughout the PSM (Morimoto et al., 2005; Shifley et al., 2008). Notch1 and Dll1 RNA expression are similar in wild-type and mutant mice (Fig. 2D), so the reduction in Notch activity observed in the posterior PSM is unlikely to be secondary to reduced expression of other Notch pathway components. The finding of distinct Notch pathway phenotypes in LfngRLFNG/+ embryos and Lfng null embryos supports the idea that LfngRLFNG is not a loss-of-function allele. Further, the different effects on Notch activity in the anterior and posterior PSM raise the intriguing possibility that RLFNG protein activity has context-dependent effects on Notch signaling.
Somitic structures are caudalized in LfngRLFNG/+ embryos
As suggested by the skeletal defects observed in LfngRLFNG/+ mice, we find that somitogenesis is disrupted in LfngRLFNG/+ embryos. At 10.5 dpc, borders between somites are absent or incomplete (Fig. 3Aa-d). LfngRLFNG/+ embryos produce myotome-like structures in the trunk, although these are fused and disorganized (Fig. 3Ae,f). Thus, expression of the RLFNG protein perturbs proper border formation and the production of epithelial somites during somitogenesis.
Somite formation and patterning are perturbed in LfngRLFNG/+ embryos. (A) At 10.5 dpc, wild-type embryos (a,c,e; n=10) form regular epithelial somites and myotomes. LfngRLFNG/+ embryos exhibit morphological segmentation defects (b; n=10), and epithelial somites are not observed in LfngRLFNG/+ sections (d; n=3). LfngRLFNG/+ embryos form myotome-like structures, but these are fused and disorganized (asterisks in f; n=3). (B) At 10.5 dpc, Mesp2 expression is slightly downregulated and diffuse in LfngRLFNG/+ embryos (c; n=5) as compared with wild-type embryos (a; n=5), but similar to that in Lfng null embryos (b; n=5). In mature somites, Tbx18 expression is severely reduced in LfngRLFNG/+ embryos (compare d with f; n=6 each), whereas Uncx is expressed throughout the mature somite region (compare g with i; n=6 each), suggesting the somites are caudalized. At this stage, Lfng null embryos exhibit poorly compartmentalized expression of both Uncx (h; n=4) and Tbx18 (e; n=4). (C) Trunk regions of 10.0 dpc embryos stained for Myog (a-c; n=3 each) and Uncx (d-f; n=5 each) show that LfngRLFNG/+ (c,f) and Lfng null (b,e) embryos produce somitic derivatives, although these are irregular, exhibiting fusions and bifurcations (arrows), and disorganization in the posterior trunk (lines in e,f). Pax1 expression in the sclerotome is less metameric in the lumbar region of LfngRLFNG/+ embryos (line in i; n=4) than Lfng null embryos (line in h; n=4). Embryos have been bisected for visualization with anterior at the top, and the forelimb bud (asterisk) maintained.
Somite formation and patterning are perturbed in LfngRLFNG/+ embryos. (A) At 10.5 dpc, wild-type embryos (a,c,e; n=10) form regular epithelial somites and myotomes. LfngRLFNG/+ embryos exhibit morphological segmentation defects (b; n=10), and epithelial somites are not observed in LfngRLFNG/+ sections (d; n=3). LfngRLFNG/+ embryos form myotome-like structures, but these are fused and disorganized (asterisks in f; n=3). (B) At 10.5 dpc, Mesp2 expression is slightly downregulated and diffuse in LfngRLFNG/+ embryos (c; n=5) as compared with wild-type embryos (a; n=5), but similar to that in Lfng null embryos (b; n=5). In mature somites, Tbx18 expression is severely reduced in LfngRLFNG/+ embryos (compare d with f; n=6 each), whereas Uncx is expressed throughout the mature somite region (compare g with i; n=6 each), suggesting the somites are caudalized. At this stage, Lfng null embryos exhibit poorly compartmentalized expression of both Uncx (h; n=4) and Tbx18 (e; n=4). (C) Trunk regions of 10.0 dpc embryos stained for Myog (a-c; n=3 each) and Uncx (d-f; n=5 each) show that LfngRLFNG/+ (c,f) and Lfng null (b,e) embryos produce somitic derivatives, although these are irregular, exhibiting fusions and bifurcations (arrows), and disorganization in the posterior trunk (lines in e,f). Pax1 expression in the sclerotome is less metameric in the lumbar region of LfngRLFNG/+ embryos (line in i; n=4) than Lfng null embryos (line in h; n=4). Embryos have been bisected for visualization with anterior at the top, and the forelimb bud (asterisk) maintained.
The skeletal phenotypes of LfngRLFNG/+ mice suggest that they exhibit somite patterning defects distinct from those observed in Lfng null embryos. Mesp2 expression, marking the presumptive rostral somite compartment in the anterior PSM (Saga et al., 1997), is expressed (although at slightly reduced levels) in LfngRLFNG/+ embryos, suggesting that rostral identity is initially specified in the PSM (Fig. 3Ba-c). However, Tbx18, which marks the mature rostral somite compartment, is severely downregulated in LfngRLFNG/+embryos (Fig. 3Bd-f), whereas the expression of a caudal somite marker, Uncx, is expanded in LfngRLFNG/+embryos (Fig. 3Bg-i). This suggests that mature somitic structures are caudalized in LfngRLFNG/+ embryos, in contrast to Lfng null embryos, which, as previously reported (Oginuma et al., 2010), exhibit a pattern of intermingled rostral and caudal cells (Fig. 3Be,h). In the trunk regions of LfngRLFNG/+ embryos, dermomyotomal (Fig. 3Ca-c) and sclerotomal (Fig. 3Cd-i) derivatives are observed, although these are irregular and exhibit bifurcations and fusions, similar to those observed in Lfng null embryos. The finding that the dominant RLFNG-related patterning defects are distinct from those observed in Lfng null embryos again supports the idea that the LfngRLFNG mutation acts through a gain-of-function.
Hes7 transcription and transcript stability are affected in LfngRLFNG/+ embryos
To examine the effects of the LfngRLFNG allele on clock function, we examined Hes7 expression in LfngRLFNG/+ embryos. Using a cDNA probe that detects processed, mature Hes7 mRNA, we observe that Hes7 mRNA levels oscillate in the PSM of wild-type embryos, whereas in LfngRLFNG/+ embryos we observe ubiquitous expression of Hes7 mRNA in the posterior PSM (Fig. 4Aa-d). To distinguish transcriptional and post-transcriptional effects on Hes7 expression in LfngRLFNG/+ embryos we examined Hes7 expression using an intron-specific probe, allowing specific detection of newly transcribed Hes7 RNAs that have not yet undergone splicing. Oscillatory transcription of Hes7 was observed in wild-type embryos, but the levels of newly transcribed Hes7 RNA in LfngRLFNG/+ embryos were significantly reduced and were restricted to the posterior PSM in all embryos examined (Fig. 4Ae-h).
Hes7 expression is perturbed and transcripts are stabilized in LfngRLFNG/+ embryos. (A) Mature Hes7 mRNA transcripts are observed throughout the PSM of LfngRLFNG/+ embryos (d; n=8), whereas levels cycle in wild-type embryos (a-c; n=4/11 phase 1, 3/11 phase 2, 4/11 phase 3). However, nascent Hes7 transcripts, revealed by an intron-specific probe, are reduced in the PSM of LfngRLFNG/+ embryos (h; n=7), whereas they are cyclic in wild-type embryos (e-g; n=2/8 phase 1, 2/8 phase 2, 4/8 phase 3). The distributions of Hes7 expression patterns were significantly different between wild-type and mutant embryos by Fisher's exact analysis (P<0.02). (B) Our observations suggest that in mutant embryos Hes7 transcription (orange) is confined to the caudal PSM, but stabilized mature transcripts persist as cells move anterior in the PSM (green).
Hes7 expression is perturbed and transcripts are stabilized in LfngRLFNG/+ embryos. (A) Mature Hes7 mRNA transcripts are observed throughout the PSM of LfngRLFNG/+ embryos (d; n=8), whereas levels cycle in wild-type embryos (a-c; n=4/11 phase 1, 3/11 phase 2, 4/11 phase 3). However, nascent Hes7 transcripts, revealed by an intron-specific probe, are reduced in the PSM of LfngRLFNG/+ embryos (h; n=7), whereas they are cyclic in wild-type embryos (e-g; n=2/8 phase 1, 2/8 phase 2, 4/8 phase 3). The distributions of Hes7 expression patterns were significantly different between wild-type and mutant embryos by Fisher's exact analysis (P<0.02). (B) Our observations suggest that in mutant embryos Hes7 transcription (orange) is confined to the caudal PSM, but stabilized mature transcripts persist as cells move anterior in the PSM (green).
The restriction of Hes7 transcription to the most caudal PSM in LfngRLFNG/+ embryos is consistent with the effects of reduced Notch activity on Hes7 (Niwa et al., 2007), and thus might reflect the reductions in NICD we observe in the posterior PSM of mutant embryos. However, our data also suggest a secondary effect on Hes7 mRNA stability or turnover, such that the Hes7 transcripts produced when cells are in the tailbud are stable enough to persist as cells progress into the anterior PSM, producing the observed stable gradient of mature Hes7 transcript in the PSM (Fig. 4B). This finding suggests the existence of previously unappreciated links between clock function and the post-transcriptional regulation of clock components.
Other clock-linked genes are affected in LfngRLFNG/+ embryos
In mouse embryos, targets of the FGF and WNT pathways have been observed to oscillate in the PSM (Dequeant et al., 2006; Wright et al., 2009), and it is clear that both FGF and WNT signaling are important for normal clock function (Aulehla and Pourquie, 2010). To determine whether RLFNG expression perturbs clock-linked genes in other pathways, we examined the expression patterns of Axin2, Spry2 and Snai1. In wild-type embryos, we observe oscillatory expression of Spry2; however, in all LfngRLFNG/+ embryos, stable Spry2 expression was observed throughout the PSM (Fig. 5A-C). Axin2 oscillations are evident in wild-type embryos, but we observed that Axin2 expression in LfngRLFNG/+ embryos was limited to an anterior stripe and faint staining in the posterior tailbud (Fig. 5D-F). Finally, whereas oscillatory Snai1 expression was observed in wild-type embryos, Snai1 was expressed stably in the posterior PSM in all LfngRLFNG/+ embryos (Fig. 5G-I). Thus, the disruption of clock activity in LfngRLFNG/+ embryos prevents the normal cyclic expression of both WNT and FGF pathway members in the PSM.
Clock-linked gene expression is perturbed in LfngRLFNG/+ embryos. Spry2 expression oscillates in wild-type embryos at 10.5 dpc (A,B; n=8/18 phase 1, 10/18 phase 2), but is slightly reduced and non-cyclic in LfngRLFNG/+ embryos (C; n=9/9). Axin2 expression oscillates in wild-type embryos at 10.5 dpc (D,E; n=5/11 phase 1, 6/11 phase 2), but is reduced in LfngRLFNG/+ embryos (F; n=8/8). Snai1 expression oscillates in wild-type embryos at 10.5 dpc (G,H; n=6/10 phase 1, 4/10 phase 2), but is noncyclic in LfngRLFNG/+ embryos (I; n=5/5). White bars delineate regions of staining. The distributions of expression patterns were significantly different between wild-type and mutant embryos by Fisher's exact analysis (P<0.02).
Clock-linked gene expression is perturbed in LfngRLFNG/+ embryos. Spry2 expression oscillates in wild-type embryos at 10.5 dpc (A,B; n=8/18 phase 1, 10/18 phase 2), but is slightly reduced and non-cyclic in LfngRLFNG/+ embryos (C; n=9/9). Axin2 expression oscillates in wild-type embryos at 10.5 dpc (D,E; n=5/11 phase 1, 6/11 phase 2), but is reduced in LfngRLFNG/+ embryos (F; n=8/8). Snai1 expression oscillates in wild-type embryos at 10.5 dpc (G,H; n=6/10 phase 1, 4/10 phase 2), but is noncyclic in LfngRLFNG/+ embryos (I; n=5/5). White bars delineate regions of staining. The distributions of expression patterns were significantly different between wild-type and mutant embryos by Fisher's exact analysis (P<0.02).
DISCUSSION
We hypothesized that secretion of LFNG from the cell provides an essential level of post-translational control in the PSM, facilitating rapid clearance of LFNG protein from cells after Lfng transcription is downregulated. This clearance could be important both in the clock, where LFNG protein levels oscillate to synchronize neighboring cells, and in somite patterning, where LFNG protein is rapidly cleared from the presumptive caudal somite compartment. To test this model, we introduced a novel mutation into the endogenous Lfng locus, replacing the N-terminal LFNG domain with sequences encoding the signal sequence and type II transmembrane domain of RFNG, creating a fusion protein called RLFNG. Previously published work demonstrates that, in tissue culture, this fusion protein is localized to the Golgi, is not secreted, and has the enzymatic function of LFNG. Thus, we predict that in vivo this allele will reduce or prevent protein processing and secretion, producing a Golgi-resident protein with the enzymatic activity of LFNG, under control of the endogenous Lfng regulatory regions. The LfngRLFNG allele in mice is associated with dominant phenotypes that are distinct from those observed in mice with loss of Lfng function, consistent with our hypothesis that LFNG processing/secretion act to terminate the function of LFNG in the cell, and that the prevention of this post-translational control will produce a hypermorphic protein.
Overall, our data are consistent with a model in which RLFNG protein is not secreted during the ‘off' phase of the clock cycle, but is stably maintained in cells after transcription ceases. This would replicate the protein behavior previously observed in tissue culture, where RLFNG protein was localized to the Golgi and not secreted into the medium (Shifley and Cole, 2008). Unfortunately, we have been unable to detect LFNG protein in mutant embryos and thus we cannot provide formal proof that the RLFNG protein in embryos is localized to the Golgi and that its levels do not oscillate. We have not identified anti-LFNG antibodies of sufficient quality to detect either wild-type LFNG or RLFNG protein in embryos, preventing direct comparison of their cellular localization or levels in vivo. Others have reported that when wild-type tagged LFNG proteins are ubiquitously overexpressed in the PSM they perturb somitogenesis, even though the LFNG protein expressed from these transgenes is not detectable (Serth et al., 2003,, 2015); thus, it is not unreasonable that RLFNG protein could be expressed at levels sufficient to disrupt segmentation without being detectable by current reagents.
Despite this caveat, our previous description of the RLFNG protein (Shifley and Cole, 2008) and the dominant nature of the phenotypes observed here provide strong support for the idea that the LfngRLFNG allele represents a gain-of-function mutation. If the mutation interfered with protein folding or activity then we would predict that it would phenocopy an Lfng loss-of-function and be recessive. By contrast, we observe dominant phenotypes in heterozygous animals that do not recapitulate the Lfng null phenotypes, and contrast with the completely normal phenotypes observed in Lfng+/− animals (Evrard et al., 1998; Zhang and Gridley, 1998). We cannot formally rule out the idea that the RLFNG protein is mistargeted in the cell or that our phenotypes arise from excessive glycosylation of its targets; however, even if these mechanisms do contribute to the observed phenotypes, this would support the idea that post-translational regulation of LFNG protein activity is important.
As predicted by our model, we observe that the feedback loops that govern the clock result in reduced transcription of the endogenous Lfng and Rlfng locus in mutant embryos. We predict that this low level of Rlfng transcript produces RLFNG protein, which is more stable than wild-type LFNG, and thus we propose that RLFNG protein is found throughout the PSM of LfngRLFNG/+ embryos (Fig. 6A). We cannot formally exclude the possibility that the LfngRLFNG allele might have some effect on transcript stability, which could exacerbate the reduced RNA expression we observe. There is no evidence that the transcript regions altered in our mutant mice harbor any RNA stability signals, and the Lfng 3′UTR, which we and others have suggested influences transcript stability (Chen et al., 2005; Nitanda et al., 2014; Riley et al., 2013), is intact in the LfngRLFNG allele. Further, our qRT-PCR analyses indicate that both alleles are expressed to similar levels in heterozygous embryos, and reductions in transcript stability could not explain the dominant nature of the phenotypes.
A model of perturbed Notch signaling in LfngRLFNG mutant embryos. (A) In wild-type PSM (left), both Lfng mRNA (top) and LFNG protein (bottom) oscillate in the posterior PSM and are stably expressed in the anterior PSM in caudal presomites. In LfngRLFNG/+ embryos (right), low levels of mRNA expression in the caudal PSM and stable expression in the anterior PSM result in stable RLFNG protein perduring throughout the PSM. (B) In the caudal PSM of LfngRLFNG/+ embryos (right), low levels of Rlfng mRNA give rise to a stable RLFNG protein, which, in the presence of DLL3, constitutively represses NOTCH1, resulting in decreased transcription of Hes7 and Lfng. In addition, increased LFNG activity inhibits Hes7 mRNA turnover through an unknown mechanism. (C) In the patterning region of the anterior PSM, LFNG protein in wild-type embryos (left) is rapidly cleared from the future caudal compartment, although Lfng transcription in the rostral compartment is maintained. Notch signaling defines the caudal somite compartment, while in the future rostral compartment MESP2 reduces Notch signaling by destabilizing Mastermind-like proteins. In LfngRLFNG/+ embryos (right), RLFNG protein is not cleared from the future caudal compartment and may perdure into mature somites. In the absence of DLL3, which is downregulated in this region, LFNG may now potentiate interactions between DLL1 and NOTCH1, increasing the levels or extent of NOTCH1 activity and caudalizing future somites by upregulating as yet unidentified Notch target genes. Since early expression of Mesp2 relies on Notch activity, the low levels of overall Notch signaling may contribute to reduced expression of Mesp2, which may further reduce the specification of the rostral compartment, enhancing the caudalization phenotype.
A model of perturbed Notch signaling in LfngRLFNG mutant embryos. (A) In wild-type PSM (left), both Lfng mRNA (top) and LFNG protein (bottom) oscillate in the posterior PSM and are stably expressed in the anterior PSM in caudal presomites. In LfngRLFNG/+ embryos (right), low levels of mRNA expression in the caudal PSM and stable expression in the anterior PSM result in stable RLFNG protein perduring throughout the PSM. (B) In the caudal PSM of LfngRLFNG/+ embryos (right), low levels of Rlfng mRNA give rise to a stable RLFNG protein, which, in the presence of DLL3, constitutively represses NOTCH1, resulting in decreased transcription of Hes7 and Lfng. In addition, increased LFNG activity inhibits Hes7 mRNA turnover through an unknown mechanism. (C) In the patterning region of the anterior PSM, LFNG protein in wild-type embryos (left) is rapidly cleared from the future caudal compartment, although Lfng transcription in the rostral compartment is maintained. Notch signaling defines the caudal somite compartment, while in the future rostral compartment MESP2 reduces Notch signaling by destabilizing Mastermind-like proteins. In LfngRLFNG/+ embryos (right), RLFNG protein is not cleared from the future caudal compartment and may perdure into mature somites. In the absence of DLL3, which is downregulated in this region, LFNG may now potentiate interactions between DLL1 and NOTCH1, increasing the levels or extent of NOTCH1 activity and caudalizing future somites by upregulating as yet unidentified Notch target genes. Since early expression of Mesp2 relies on Notch activity, the low levels of overall Notch signaling may contribute to reduced expression of Mesp2, which may further reduce the specification of the rostral compartment, enhancing the caudalization phenotype.
Our data suggest that Rlfng expression has context-dependent effects on Notch activation. The reduction of NICD in the posterior PSM of mutant embryos is consistent with the idea that LFNG represses NOTCH1 activation in the clock, and that expression of RLFNG protein results in ubiquitous (but not complete) inhibition of Notch activity in this region. However, activation of NOTCH1 in the anterior PSM of LfngRLFNG/+ embryos is not as severely inhibited by RLFNG, appearing as a broad band that is not refined into the caudal somite compartment seen in wild-type embryos (Fig. 2Bh,i). The phenotypes of mutant embryos support the idea that Notch signaling is affected in different ways, with clock phenotypes suggestive of reduced Notch activity, and pattering phenotypes suggestive of expanded or ectopic Notch activity.
In the clock, LFNG has been proposed to act in the signal-sending cell to inhibit Notch activation through an unknown mechanism that is enhanced by co-expression of DLL3 (Okubo et al., 2012). We suggest that expression of the mutant allele produces RLFNG protein that is tethered in the Golgi, perturbing normal clock function by altering the coordinated, cyclic activation of Notch in the PSM and dysregulating the feedback loops that regulate this activation (Fig. 6B). In this model, RLFNG activity in posterior PSM cells would work with DLL3 to inhibit activation of NOTCH1 by DLL1 in neighboring cells. This reduction in Notch activity would in turn affect the transcription of Hes7, Lfng, and other clock-linked genes. The segmentation clock phenotypes observed in LfngRLFNG mice represent the steady-state expression of clock genes subsequent to the loss of clock synchrony. The LfngRLFNG mutant embryos described here exhibit a dramatic reduction in the level of NOTCH1 activation and in the transcription of Notch targets such as Hes7 and Nrarp. The crosstalk mechanisms that coordinate Notch, WNT and FGF oscillations in the clock are complex, and further work will be necessary to fully understand how RLFNG protein affects other clock activities. However, these results support our initial hypothesis that secretion of LFNG from the cell acts to terminate its activity in the segmentation clock.
In contrast to the effects in the posterior PSM, in the anterior PSM and mature somites the phenotypes observed in LfngRLFNG/+ embryos are suggestive of increased or expanded Notch activation, with loss of rostral markers and expansion of caudal markers (Fig. 3). Dorsal rib tissues arise from the caudal somite (Aoyama and Asamoto, 2000); thus, the somite caudalization observed in LfngRLFNG/+ embryos is consistent with the severe dorsal rib fusions observed in adult mutant animals (Fig. 1), which are similar to those seen in Tbx18 mutant mice or mice with a hypomorphic Mesp2 allele (Bussen et al., 2004; Nomura-Kitabayashi et al., 2002). Taken together, these phenotypes resemble those seen when ligand-independent NICD was overexpressed in the PSM (Feller et al., 2008) or when Notch activity is exogenously activated in the rostral somite (Sasaki et al., 2011) or derepressed in the rostral somite due to loss of Mesp2 (Takahashi et al., 2013). Thus, these phenotypes suggest that the protein expressed from the LfngRLFNG allele potentiates or expands the regions of NOTCH1 activation in the presomites of mutant mice. The resulting expansion of Notch target genes would be predicted to caudalize somitic structures.
In the anterior PSM of wild-type embryos, DLL1 is upregulated in the presumptive caudal compartment and DLL3 is confined to the presumptive rostral compartment, where it is co-expressed with wild-type LFNG. Notch activation (visualized via NICD) is apparent in stripes in the anterior PSM and in the caudal compartments of mature somites. In LfngRLFNG mutant embryos, we observe a single broad band of NICD in the anterior PSM, and low but ubiquitous levels of NICD in the mature somitic region (Fig. 2B). The mechanism(s) that perturb Notch activation in the anterior PSM of LfngRLFNG embryos are not clear, but our data are consistent with the idea that, in LfngRLFNG/+ embryos, RLFNG protein would be more stable than wild-type LFNG and thus would persist in the caudal presomite or mature somite after Rlfng gene expression is downregulated. This is in contrast to wild-type LFNG protein, which would be rapidly cleared out of these structures, presumably by SPC6 (also known as PCSK5), which we have shown can process LFNG protein and is highly expressed in the anterior PSM (Shifley and Cole, 2008).
In the absence of DLL3, this residual RLFNG protein in the presumptive caudal somite compartment could potentiate activation of NOTCH1 by DLL1, as has been shown in other systems (Hicks et al., 2000; Kato et al., 2010; Yang et al., 2005) (Fig. 6C), promoting the caudalized phenotypes that we observe. This expansion of Notch activity could be increased by the observed reduction of Mesp2 in mutant embryos, as MESP2 protein in wild-type embryos directly reduces Notch activity in the future rostral compartment by destabilizing Mastermind-like protein (Sasaki et al., 2011). Finally, we predict that RLFNG protein will persist in the mature somite region where wild-type Lfng is not expressed, and it is not known what effects it might have there. We have previously shown that SPC6, which can cleave LFNG, is highly expressed in the rostral component of somite S–1, which might suggest that clearance of LFNG protein from this region is important (Shifley et al., 2008). A conditionally activatable LfngRLFNG allele or an allele targeting RLFNG to the anterior PSM might help address these questions in the future. It is difficult to ascribe specific aspects of the skeletal phenotype to perturbations in clock function as opposed to alterations in somite patterning, as loss of clock function may influence patterning defects. However, the somite caudalization and the loss of metameric Pax1 expression observed in LfngRLFNG/+ embryos are similar to the outcomes observed when inappropriate Notch activation is driven in the rostral somite compartment (Sasaki et al., 2011; Takahashi et al., 2013); thus, we suggest that some of the observed phenotypes are related to altered patterning.
It is not clear what, if any, role secreted LFNG plays during somitogenesis. Mathematical models suggesting that secreted LFNG functions in synchronizing the segmentation clock have been proposed (Cinquin, 2003), but the phenotypes reported here are unlikely to be caused solely by a loss of LFNG secretion. LfngRLFNG/+ mice retain a wild-type copy of Lfng that could produce the secreted form, although at low levels due to the observed reductions in Lfng transcripts. Further, the skeletal and patterning phenotypes, as well as the effects on NICD levels, in LfngRLFNG/+ mice are distinct from those observed in Lfng null mice, providing little support for the idea that the phenotypes observed in LfngRLFNG/+ mice are due to any loss of LFNG activity. The finding that the LfngRLFNG allele does not phenocopy the effects of loss of LFNG, together with the dominant nature of these phenotypes, support the idea that at least some of the phenotypes observed in our mutant mice are the consequence of a perturbation of an intracellular function of LFNG. However, our results do not formally rule out the possibility that the LFNG protein has an extracellular function, which we expect would also be perturbed in our mutant mice.
Finally, our work suggests a link between clock activity and the post-transcriptional regulation of clock components. Expression of RLFNG has effects on both transcriptional and post-transcriptional regulation of Hes7. In LfngRLFNG/+ embryos, NICD levels are reduced in the posterior PSM and Hes7 transcription is confined to the posterior PSM. This is consistent with previous findings suggesting that Notch activity is not required for the initiation of Hes7 transcription in the most posterior PSM (Niwa et al., 2007). However, mature Hes7 transcripts appear to be stabilized in the PSM of LfngRLFNG/+ embryos. This represents the first mouse mutation known to affect Hes7 mRNA turnover, although mutations that affect RNA stability have been observed for Hes homologs in Xenopus and zebrafish (Davis et al., 2001; Dill and Amacher, 2005). Thus, these findings suggest that the rapid turnover of Hes7 RNA in the mouse PSM may be regulated by an unknown mechanism that is linked to the segmentation clock or to the Notch signaling pathway. Further analysis of the mice reported here might shed light on this important question.
MATERIALS AND METHODS
Targeted mutation and genotyping
Lfng sequences from the NotI-AatII sites in exon 1 were replaced with the RFNG N-terminus amplified with primers (5′-3′) ATGCGGCCGGCGGCCACCATGAGCCGTGCGCGGCGG and GGTTCTTCCGAGTGGTCTTG, replacing the LFNG pre/pro region with the signal sequence of RFNG fused at LFNG D113. The 5′ arm contains a floxed Neo-testis Cre cassette, which is excised upon passage through the male germline (Bunting et al., 1999). LfngRLFNG mice were genotyped with primers SC286 (TTGGGTCTATCTGGGAAACG) and SC287 (GCGACTCATCCAGACACAGA) producing a 149 bp wild-type band and a 250 bp mutant band. LfngtmRjo1 (Lfng null) mice were genotyped as described (Shifley et al., 2008). Mice were maintained on a mixed 129/Sv×FVB/J background in an specific-pathogen-free facility under the care of OSU Laboratory Animal Resources. All protocols were approved by the OSU IACUC.
Skeletal preparations and histology
Embryos were harvested at 17.5 dpc. Alcian Blue/Alizarin Red staining was performed as described (McLeod, 1980). For histology, embryos were fixed in Bouin's fixative, processed for paraffin sections, and stained with Hematoxylin and Eosin.
Whole-mount RNA in situ hybridization and immunohistochemistry analysis
Embryos were collected from timed pregnancies (noon of day of plug detection is 0.5 dpc) and fixed in paraformaldehyde. Hybridization with digoxigenin-labeled probes was performed as described (probes are listed in Table S1). Whole-mount immunohistochemistry to detect NICD (Cell Signaling 4147, 1:500; Table S1) was performed as previously described (Shifley et al., 2008).
RT-PCR
The posterior PSM of individual 10.5 dpc embryos was collected in cold PBS. Endogenous Lfng RNA levels were assessed using gene-specific TaqMan assays (Applied Biosystems Mm00456128_m1*). Lfng values were normalized to Gapdh, and the highest level of wild-type expression was set to 100 for comparison across embryos. For allele-specific RT-PCR, cDNA was produced from individual embryos (SuperScript III, Invitrogen). Amplification of Lfng alleles was performed with primers SC1019 (CTGGCTGTGTTGCTGCTACT), SC-1020 (CCCATCAGTGAAGATGAACG) and SC1021 (GATCCCGGAGTCCTCACC) producing a 237-bp LfngRLFNG-specific band and a 267 bp wild-type-specific band. In conventional PCR, all three Lfng primers were included, and PCRs were run for 35 cycles to obtain robust signals for both wild-type and mutant embryo samples. For qRT-PCR, levels of Lfng or Rlfng were assessed using an Applied Biosystems StepOne Plus machine and SYBR Select Master Mix. Hprt was amplified using primers AGCTACTGTAATGATCAGTCAACG and AGAGGTCCTTTTCACCAGCA. Absolute quantification was performed using standard curves of linearized plasmids, followed by normalization to Hprt.
Statistical analyses
For oscillatory expression patterns, Fisher's exact analyses were used to compare the distributions of expression patterns among the oscillatory phases between wild-type and mutant embryos. Generally, the invariant mutant expression pattern was assigned to the oscillation phase it most closely resembled for these analyses (i.e. the expression of Lfng of all mutant embryos was assigned to ‘phase 3'). In the case of Spry2, as the mutant expression pattern was distinct from any wild-type pattern, a third category of expression was assigned.
Acknowledgements
We thank S. Amacher and members of the S.E.C. laboratory for comments.
Author contributions
S.E.C. developed concepts/approach, performed some experiments, and contributed to data analysis and manuscript preparation. D.R.W. and E.T.S. designed and performed most experiments, and contributed to data analysis and manuscript preparation. K.M.B. performed and assisted with the design of some experiments and contributed to manuscript editing.
Funding
This work was supported by a grant from the National Science Foundation [# IOS-0919649 to S.E.C.].
References
Competing interests
The authors declare no competing or financial interests.