Knockdown of CRIM1 in Xenopus embryos causes defects in neuronal structures development
Using available chick CRIM1 sequence (accession #NM_204425) to design primer sets, we PCR amplified cDNA products from a stage 28 Xenopus laevis cDNA library and identified two distinct sequences that had extensive homology to chick CRIM1. Based on the high degree of homology, these clone families represented the Xenopus laevis A and B genes. We used this sequence information (, accession number pending) to design PCR primers, antisense Morpholino oligonucleotides (MO, , ) and in situ hybridization probes for Xenopus laevis CRIM1.
CRIM1 is expressed in the neural plate and is required for development of neural structures.
Morpholino oligonucleotides used in this study.
RT-PCR analysis for CRIM1 on a staged series of embryos () showed that CRIM1 mRNA is detected in the early neurula at stage 12. For comparison, n-tubulin mRNA was detected in the late neurula at stage 22. By in situ hybridization, CRIM1 was detected in the neural plate of stage 12.5 embryos (). Expression of CRIM1 in neural structures continued and at stage 18, albeit faintly detected, in posterior neural tube () as well as anterior neural structures including optic vesicles (). CRIM1 expression at stage 22 was detected in the early somites and weakly in neural structures (). The hindbrain, cement gland and somites were all locations of CRIM1 expression at stage 35 ().
CRIM1 loss-of-function experiments in Xenopus laevis
were performed using antisense MO-mediated translation and splicing blocking 
. Sequence differences in the 5′ untranslated region of the CRIM1
A and B genes () required that we use a mixture of MOs (XLCAB, ) for translation blocking. To design splicing-blocking MOs, we first identified Xenopus tropicalis
sequences in the available database (JGI Genome Browser) and used that sequence to PCR amplify and sequence Xenopus laevis
genomic clones. The CRIM1
A and B genes also had sequence changes in the exon 2 splice donor region () that necessitated a mix of MOs (XLCSDAB, ) to target both mRNAs. Using two sets of PCR primers that detected either unspliced or spliced mRNA () we confirmed that the MOs targeted to the splice donor of CRIM1
exon 2 suppressed splicing.
Translation and splicing blocking CRIM1 MOs injected into a dorsal blastomere at the 4-cell stage produced dramatic effects on the development of neural structures. In a typical experiment where 15 ng each of XLCA and B were injected, more than 70% of embryos had major defects including a small or missing eye on the injected side ( and ). Tracing of MO distribution with coinjected Dextran Alexa488 confirmed that the affected region of the embryo received MO but that any remaining neural tube was tracer negative (). Histological assessment of affected Xenopus embryos at stage 42 confirmed the neural tube and eye were both missing on the injected side (data not shown). In embryos injected bi-laterally at the 2-cell stage with 30 ng each XLCA and B MOs, a loss of anterior neural and head structures resulted but ventral and posterior structures were retained (). This phenotype induced by loss of CRIM1 in the whole embryo by administering the MOs at this stage correlates well given the expression pattern of CRIM1 in the developing neural plate.
Phenotype summary for CRIM1 Morpholino injections.
Since the absence of an eye served as a simple read-out for phenotype severity, we assessed changes from MOs injection in different amounts. There was a dose response for both translation and splicing blocking MOs and that each produced the same phenotype (). Injection of ventral blastomeres with the translation blocking XLCAB combination had a minimal effect (, vent). MOs () in which 5 of the nucleotides were mismatched had a greatly reduced effect though this was not zero (). Since 5 nucleotide mismatch MOs are known to retain some activity at the concentrations used here 
we also used the standard control MO (GeneTools) that has no measurable activity as a control and observed no obvious phenotype ().
We also determined whether co-injection of a MO-resistant CRIM1 mRNA with the XLCAB MOs resulted in phenotypic rescue. Though we did not observe a complete reversal of the effects of the XLCAB MOs, the MO-resistant xCRIM1 mRNA reduced the percentage of embryos showing small or missing eyes (). Together, the activity of both MO types in producing the same phenotype, the correlation of that phenotype with the expression domain of CRIM1, suppression of CRIM1 mRNA splicing with XLCSDAB MOs and a degree of phenotype rescue with xCRIM1 expression suggest the antisense oligonucleotides are specific.
The absence of neural structures in tailbud stage embryos was consistent with the loss of neural plate integrity at earlier stages. Examination of XLCAB MOs injected pigmented embryos at stage 15 when neural plate morphogenesis is occurring revealed that the injected side had defects in neural plate formation. Specifically, failure of the neural plate boundary (the neural folds) to move toward the midline produced embryos with a pronounced asymmetry (). In many XLCAB MOs injected embryos, cells were seen sloughing from the surface of the injected side (, red arrowheads). In a typical experiment using 15 ng each XLCA and XLCB (), 20/70 embryos (28.6%) show severe cell sloughing. Time-lapse video microscopy of embryos bi-laterally injected with MOs (dorsal blastomeres, XLCAB at the 4-cell stage) in some cases showed a mild phenotype of delayed neural fold morphogenesis with a failure of anterior neural tube closure (Video S1
) and in others a severe failure of cell adhesion across the entire neural plate (Video S2
). This suggested that CRIM1 might have an essential role in promoting cell adhesion or suppressing cell death within the neural plate.
Disruption of neural fold morphogenesis in CRIM1 MO injected embryos.
Reduced cadherin junctional complexes is a primary consequence of CRIM1 loss-of-function
To distinguish between these two possibilities, we first determined whether the level or localization of cadherins that are critical adhesion molecules in Xenopus
neural plate 
might be affected in CRIM1
knockdown embryos. We coinjected XLCAB MOs with a tracer mRNA encoding GFP at the 4-cell stage and then performed whole-mount immunolabeling for cadherins at stage 13 (early neurula stage). In these preparations, an apical cadherin junctional complex is identified revealing patterns of cell packing and cell size at the surface (). In this case we controlled the experiment by injecting the GFP tracer mRNA alone. In other experiments co-injecting control MOs with dextran tracer gave identical results (, , ). In control embryos we see slight junction-to-junction variation in labeling intensity for both E-cadherin () and C-cadherin (), but this did not correlate with GFP expression.
CRIM1 is required for junctional localization of E- and C-cadherin in the neural plate.
Cadherin junctional complex deficiency is the primary response to CRIM1 loss-of-function.
CRIM1 is required for junctional localization of ß-catenin in the neural plate.
Restoration of CRIM1 expression can rescue ß-catenin level in CRIM1 loss-of-function embryos.
By contrast, when the GFP mRNA and the CRIM1 MO were co-injected there were dramatic changes in cadherin labeling in GFP expressing cells. At low magnification () tracer positive regions have reduced immunoreactivity for both E-cadherin and C-cadherin. Higher magnification () shows the precise correlation between GFP expression and reduced junctional labeling intensity. In addition, a junction between two tracer positive cells generally has a low level of cadherin immunoreactivity compared with junctions between a tracer positive and a tracer negative pair or between two tracer negative cell junctions (). To quantify the E- and C-cadherin labeling, we measured pixel intensities over a curved line interval superimposed along junctional labeling between two cells. When normalized to the value of junctions between pairs of tracer-negative cells, a tracer positive-tracer negative pair showed no reduction in labeling intensity whereas tracer-positive pairs showed significantly reduced labeling intensity for E-cadherin () and C-cadherin (). At higher magnification, some tracer positive cells have a rounded shape and a greater apical surface area than their tracer-negative neighbors () disrupting the pattern of cell packing.
While these changes in junctional cadherin levels and cell shape were consistent with a role for CRIM1 in adhesion, it remained possible that the cells with low cadherin levels were undergoing apoptosis as a primary response to CRIM1 loss-of-function. To determine whether this occurs, we performed two different assays for cell death. First, we injected embryos with either the fluorescent dextran tracer alone or with tracer plus 15 ng each XLCA and B MOs into a dorsal blastomere at the 4-cell stage. We harvested embryos at stage 13, permeablized and performed whole-mount TUNEL labeling (Fig. S1
). As a positive control, we used the same combination of control and MO-injected embryos but treated them with DNase I to nick genomic DNA and enhance TUNEL labeling (Fig. S1
). DNase I-treated embryos were TUNEL labeled; control or XLCAB-injected embryos without DNase-I treatment were not. Embryos were injected with the same amount of MOs that reliably caused reduced junctional cadherin labeling at the same analyzed stage ().
Since it can be argued that TUNEL labeling monitors a late event in the activation of cell death pathways, we also performed labeling for activated Caspase 3, an early marker of cell death pathway activation combined with labeling for C-cadherin (). In this set of experiments, we analyzed CRIM1 knockdown embryos that showed a patch of de-adhering cells judged morphologically (). We performed quantification of pixel intensity for the dextran tracer, C-cadherin and activated Caspase 3 along 450 pixel line intervals extending through tracer-negative to tracer-positive regions (). These data are graphically represented in pixel intensity histograms (). Regions of the micrograph containing the line interval are reproduced at higher magnification below the histogram (). We analyzed 14 examples each of control MO and XLCAB-injected embryos and found consistent results.
In embryos co-injected with the tracer and the standard control MO, lineage tracer-positive cells retained strong C-cadherin junctional staining (, middle panel and 4C, red). Activated caspase-3 levels, with the exception of the occasional positive cells (, blue arrowhead), were consistently low across the whole embryo (, right panel) and along the line interval used for analysis (, blue). By contrast, in embryos co-injected with the tracer and 15 ng each XLCAB MO, C-cadherin labeling was consistently lower in tracer-positive regions as seen in the micrographs (, middle panel, 4F, Ccad) and also when comparing red channel pixel intensities in tracer-negative and positive regions on the histogram ().
We performed quantification of these signals by measuring pixel intensities over 150 pixel line intervals located exclusively in tracer negative (control MO and XLCAB injected embryos), tracer positive, adherent (control MO and XLCAB injected embryos), or tracer positive, non-adherent regions (XLCAB injected embryos only) in 8 different embryos. In XLCAB injected embryos, C-cadherin labeling was significantly reduced in both adherent and non-adherent tracer-positive regions compared with tracer-negative regions (, red bars. A number lower than 1 indicates reduction of C-cadherin expression in MO injected regions). Importantly, adherent, MOs injected regions with reduced C-cadherin levels show no change in the level of activated Caspase 3 (, blue bars). In addition, activated caspase 3 levels only increase dramatically when cells show non-adherent morphology (, blue bars). These data argue that the primary consequence of CRIM1 loss-of-function is a diminished level of cadherin junctional complex and that cell de-adhesion followed by activation of cell death pathways is a secondary consequence.
CRIM1 is required for ß-catenin localization to junctional complexes
The cadherin junction defects apparent in CRIM1 knockdown experiments prompted us to determine whether CRIM1 might regulate the level or distribution of other major adhesion complex proteins. To assess this, we generated embryos co-injected with the dextran tracer and control or XLCAB MOs and labeled for both C-cadherin and ß-catenin. As described above, we chose to analyze experimental embryos that had regions of non-adherent cells as judged morphologically (). This analysis is illustrated and quantified as described for .
Control MO injected embryos showed levels of C-cadherin and ß-catenin signal that were consistent across tracer-negative and tracer-positive regions of the embryo (). By contrast, tracer-positive regions in XLCAB-injected embryos showed reduced levels of both C-cadherin and ß-catenin regardless of whether these regions were adherent or non-adherent (). To quantify the level of C-cadherin and ß-catenin, we generated pixel intensities over 150 pixel intervals on control MO-positive, and XLCAB-positive adherent and non-adherent regions. We then quantified the changes in average pixel intensities in MO-positive (tracer-positive) regions compare to those in MO-negative regions for both C-cadherin and ß-catenin labeling (). Compared with control MO regions, the XLCAB MO resulted in a mild but statistically significant reduction in C-cadherin signal and a more pronounced reduction in ß-catenin signal (). Interestingly, the level of C-cadherin signal reduced dramatically when cells become non-adherent while ß-catenin signal showed no further reduction (). This suggested that a primary consequence of CRIM1 loss-of-function is the failure of ß-catenin to stably associate with cadherin junctional complexes.
Compromise of the cadherin junctional complex leads to defects in apical-basal epithelial polarity 
. To determine if this feature of neural plate epithelial cells might be changed with CRIM1 loss-of-function, we performed similar experiments by co-injecting XLCAB with dextran tracer and performed immunofluorescent labeling on cross sections of the neural epithelium. Embryos displaying a mild phenotype were analyzed midway through neurulation at stage 16. The tracer was generally (, green) but not always (, green) distributed in a region that abutted the midline as would be expected for injection of a dorsal blastomere at the 4-cell stage. Tracer positive regions had a markedly different labeling pattern for ß-catenin. In unaffected neural epithelium (tracer negative, ) the neural epithelium has intense ß-catenin labeling at cell junctions and the columnar cell shape of the outermost epithelial layers is distinct (tracer negative, , grayscale panels). In all regions receiving the XLCAB MOs (, green region with dashed white line boundary) junctional ß-catenin labeling level is lower, the cells show a more rounded shape and the epithelium is disorganized. Out of 24 embryos each of experimental and control, we found polarity defects that were restricted to the tracer-positive regions in 7 experimental embryos.
We then determined whether restoration of CRIM1 expression would rescue the abnormal distribution of ß-catenin in CRIM1 knockdown cells. To this end, a MO-resistant, FLAG-tagged full-length CRIM1 mRNA (CRIM1-FL) was co-injected with XLCAB MOs into a dorsal blastomere at the 4-cell stage. The expression level of the tagged protein was measured by comparing average pixel intensities over a 150 pixel line interval placed in tracer positive and tracer negative areas (, white lines). Injection of the mRNA resulted in robust expression of tagged full-length CRIM1 with or without co-injection of XLCAB MOs ( right panels, blue bars). Whole-mount ß-catenin labeling was performed on embryos injected with different combinations of MOs and mRNA. We found that while injecting CRIM1 mRNA alone did not change the expression of ß-catenin (, middle panel; ), co-injecting CRIM1 mRNA with XLCAB MOs restored the ß-catenin intensity ( middle panels) to the normal level of ß-catenin as in embryos injected with control MO (). Combined, these data suggest CRIM1 has an essential role in stabilizing the cadherin junctions.
CRIM1 complexes with ß-catenin and N-cadherin via its cytoplasmic domain
As a first step in understanding the mechanism of action of CRIM1, we determined whether multiple CRIM1 molecules could associate in a complex. We co-expressed a FLAG-tagged ectodomain form (, top line) with a series of deletion mutants carrying C-terminal V5 tags () and determined whether this would coimmunoprecipitate (co-IP) from HEK293 cells. According to immunoblots with appropriate antibodies, all proteins expressed well (, left panels) and the V5 tagged proteins could also be efficiently IPd (, far right). Anti-V5 IP followed by immunoblot with anti-FLAG showed all deletion mutants of V5 tagged CRIM1 could form complexes with CRIM1-FL-ED (, center left). These data indicate that CRIM1 can form complexes where multiple CRIM1 molecules are present. These data also show that an N-terminal region containing the IGFBP-like domain is sufficient for formation of this complex.
CRIM1 self-associates via the N-terminal domain and forms a complex with ß-catenin and N-cadherin via the C-terminal domain.
The apparent role of CRIM1 in stabilizing cadherin junctions shown by knockdown and rescue experiments prompted us to determine whether CRIM1 might directly interact with major adhesion complex proteins. To this end we over-expressed epitope-tagged CRIM1 in HEK293 cells, and determined whether CRIM1 could be IPd in these complexes (data not shown). When anti-ß-catenin antibodies were used for IP, CRIM1 was readily detected by immunoblot (data not shown). We then generated mutant forms of CRIM1 that lacked the cytoplasmic domain (). We also used two different locations for epitope tagging given the possibility that a C-terminal epitope tag might prevent a cytoplasmic domain interaction (). All four modified CRIM1 proteins expressed well in HEK293 cells (, left panel) and could be IPd effectively with the antibody to the appropriate tag (, right panel). Only CRIM1 with an intact cytoplasmic domain would form a complex with ß-catenin (, center left) through IP using anti- ß-catenin antibodies.
To determine whether the cytoplasmic domain of CRIM1 was sufficient for ß-catenin complex formation, we expressed CRIM1-cyt (consisting of the secretory leader, transmembrane and cytoplasmic domains, ) in 293 cells and performed ß-catenin IPs. Both CRIM1-cyt and the full-length CRIM1 expressed well as indicated by an anti-V5 immunoblot of cell lysates (, left panel – tracks 3 and 4 are duplicates). Using anti-ß-catenin antibodies, both full-length CRIM1 and CRIM1-cyt co-IPd (, right panel). We used antibodies to the FLAG epitope in CRIM1-FL and CRIM1-FLΔcyt () in reciprocal IPs and detected ß-catenin () in immunoblots. In lysates from CRIM1-FL expressing cells, total ß-catenin levels appeared unchanged where a CRIM1-ß-catenin complex was demonstrated via co-IP (). These data provide strong evidence that CRIM1 and ß-catenin exist in the same complex. We could not convincingly demonstrate a direct interaction between a variety of recombinant forms of the CRIM1 cytoplasmic domain and ß-catenin in vitro (data not shown).
The CRIM1 knockdown adhesion defect, together with co-existence of CRIM1 and ß-catenin in a protein complex raised the possibility of CRIM1 association with cadherins. N-cadherin is expressed in HEK293 cells whereas E-cadherin is not (data not shown). When CRIM1 was over-expressed, anti-N-cadherin antibodies IPd CRIM1 (). Formation of a CRIM1-N-cadherin complex was also dependent upon the presence of an intact CRIM1 cytoplasmic domain (). Combined, these data indicate that CRIM1 can form complexes with ß-catenin and N-cadherin via its cytoplasmic domain. This, with reduced junctional cadherin levels in Xenopus CRIM1 knockdown expreriments, suggested that the adhesion defect resulted from disruption of cadherin-dependent junctional complexes.