Establishment of the left-right axis is essential for normal organ morphogenesis and function. Ca2+ signaling and cilia function in the zebrafish Kuppfer's Vesicle (KV) have been implicated in laterality. Here we describe an endogenous Ca2+ release event in the region of the KV precursors (dorsal forerunner cells, DFCs), prior to KV and cilia formation. Manipulation of Ca2+ release to disrupt this early flux does not impact early DFC specification, but results in altered DFC migration or cohesion in the tailbud at somite stages. This leads to disruption of KV formation followed by bilateral expression of asymmetrical genes, and randomized organ laterality. We identify β-catenin inhibition as a Ca2+-signaling target and demonstrate that localized loss of Ca2+ within the DFC region or DFC-specific activation ofβ-catenin is sufficient to alter laterality in zebrafish. We identify a previously unknown DFC-like cell population in Xenopus and demonstrate a similar Ca2+-sensitive stage. As in zebrafish,manipulation of Ca2+ release results in ectopic nuclearβ-catenin and altered laterality. Overall, our data support a conserved early Ca2+ requirement in DFC-like cell function in zebrafish and Xenopus.
INTRODUCTION
Although vertebrates appear bilaterally symmetrical from the outside, the heart, lungs, liver and gut are carefully positioned across the left-right(LR) axis, and the development of this asymmetry is highly conserved across species. Deviations from the normal internal organ LR arrangement can cause lethality and congenital heart defects (reviewed by Bisgrove et al., 2003). Embryonic organ laterality is preceded by molecular and physiological asymmetries. Asymmetrical H+/K+-ATPase expression and gap junctional communication in early cleavage stage Xenopus embryos have been proposed to modulate LR signals but may be species specific(reviewed by Levin and Palmer,2007). During somite stages, asymmetrical gene expression in the lateral plate mesoderm (LPM) becomes apparent. Of note is the left-sided LPM expression of the secreted transforming growth factor-β (TGFβ)related factor nodal, a conserved feature in all vertebrate species examined to date (reviewed by Ahmad et al.,2004). Likewise, downstream targets of nodal, including lefty1 and pitx2, have conserved asymmetric expression(reviewed by Hamada et al.,2002; Raya and Belmonte,2006; Wright and Halpern,2002). Asymmetrical nodal expression and LR axis determination in post-gastrula embryos are correlated with the proper structure and function of monocilia located in the embryonic node. Rotation of these monocilia generates a leftward fluid flow in the lumen of the node and is thought to be required for proper LR asymmetry(Hirokawa et al., 2006; Nonaka et al., 1998; Okada et al., 1999). Despite the identification of multiple LR signals, the degree of conservation, as well as the integration and coordination of these independent physiological and molecular signals, is not clearly defined.
In the zebrafish, a structure called Kuppfer's Vesicle (KV) is structurally and functionally homologous to the node, and cilia in the KV are thought to generate critical laterality signals that are propagated to the LPM. The KV is derived from a cluster of non-involuting cells called the dorsal forerunner cells (DFCs), which migrate ahead of the dorsal blastoderm during gastrulation. These cells contribute to both the tailbud and the KV during somite stages (Cooper and D'Amico,1996; Melby et al.,1996). Furthermore, the DFCs express several genes required for LR patterning and ablation of DFCs disrupts LR patterning(Amack and Yost, 2004; Essner et al., 2005).
Studies in several vertebrate model organisms implicated a requirement for Ca2+ signals in LR asymmetry, downstream of fluid flow in the node(Hirokawa et al., 2006) and/or of H+/K+ ATPase activity (reviewed by Tabin, 2006). Leftward fluid flow within the node, generated by left-right dynein containing cilia, is thought to stimulate mechanosensory cilia at the node periphery. These cilia contain polycystin-2/PKD-2, a Ca2+-permeable cation channel, and their stimulation triggers elevated intracellular Ca2+ levels at the left edge of the mouse node (McGrath et al., 2003). In zebrafish, Ca2+ fluxes near the zebrafish KV have been reported and are proposed to be required for normal LR patterning (Sarmah et al.,2005). Additionally, zebrafish embryos mutant for pkd2have disrupted expression of laterality markers and organ laterality defects(Schottenfeld et al., 2007). Although the mouse and zebrafish loss-of-function phenotypes are not identical, both models implicate a role for Ca2+ release. In an alternative mouse model, the leftward flow of vesicular particles containing sonic hedgehog (shh) has been postulated to act in a distinct atypical signaling pathway to activate Ca2+ on the left side of the node (Tanaka et al.,2005). Elevated Ca2+, in either case, is thought to act via an unknown mechanism, to induce left-sided gene expression(Brennan et al., 2002; Hashimoto et al., 2004; Marques et al., 2004).
In contrast to intracellular Ca2+ release, a role for left-sided elevation of extracellular Ca2+ has also been proposed. In the chick, differential H+/K+ ATPase activity is thought to set up a spatial gradient of extracellular Ca2+ by the end of gastrulation, which may be sensed by Notch to activate asymmetrical gene expression (Raya et al.,2004). In zebrafish, disruption of H+/K+ATPase activity also leads to LR asymmetry defects(Kawakami et al., 2005) and disrupts KV cilia number and length (Adams et al., 2006). However, in these studies, it was unclear whether intracellular Ca2+ was also increased on the left side of the node/KV. Also, in the case of H+/K+ ATPase inhibition, shh was induced ectopically, an event which could influence laterality.
Thus, several vertebrate models implicate a role for Ca2+signaling in LR asymmetry (Shimeld,2004), and the data are consistent with a function downstream of H+/K+ ATPase or node/KV cilia during somitogenesis. However, it has not been determined whether Ca2+ release also has a role in parallel or upstream to KV/node formation and function. In this study,we identified endogenous Ca2+ release in the DFC region after the onset of gastrulation, distinct from the events described above and prior to cilia and KV formation. We show that transient inhibition of this Ca2+ release results in altered expression of asymmetric markers such as southpaw, lefty1/2 and pitx2, randomized placement of organs relative to the LR axis, all preceded by the reduction/loss of KV formation. We further provide several lines of evidence to suggest that the regulation of β-catenin activity is a critical target of Ca2+in the DFCs. First, we show that inhibition of Ca2+ release in the DFC region causes a marked increase in nuclear β-catenin andβ-catenin transcriptional activity. Second, targeted activation ofβ-catenin in the DFCs is sufficient to alter laterality. Finally, we describe cells with DFC-like properties in Xenopus and establish that the epiboly/gastrula Ca2+-sensitive event is conserved in frogs, as a pulse of Ca2+ inhibition in Xenopus leads to altered organ laterality and nodal expression as well as ectopic accumulation ofβ-catenin. Thus, in addition to conserved node-like/cilia development in fish and frog, our results indicate a conserved early Ca2+requirement in LR patterning.
MATERIALS AND METHODS
Embryo manipulation
Zebrafish embryos were collected from natural spawning. Embryos were rinsed with embryo medium (Westerfield,1995) and staged according to Kimmel et al.(Kimmel et al., 1995). Live embryos were photographed after orienting in 3% methylcellulose. Xenopus eggs recovered from females were fertilized in a sperm suspension and maintained in 0.1× MMR.
Pharmacological reagents
Thapsigargin (2.5 μM; Molecular Probes), BODIPY-FL-Thapsigargin (2.5μM; Molecular Probes), Cyclopiazonic Acid (200 μM; Calbiochem) or Valproic Acid (0.2 mM; Sigma) were diluted in embryo medium. Zebrafish embryos were incubated in thapsigargin for the indicated time and then washed three times in embryo medium. Xenopus (stage 11.5) were incubated for 20 minutes in thapsigargin. Xestospongin C (1-2 μM; Calbiochem), was co-injected with Texas Red lineage marker (Molecular Probes) into the yolk of 256- to 512-cell embryos.
Morpholino antisense oligonucleotides (MO)
Control: CCTCTTACCTCAGTTACAATTTATA and Axin1: ACTCATGCTCATAGTGTCCCTGCAC MOs were purchased from Gene Tools, LLC. MO (8 ng) or in vitro transcribedβ-catenin 55C RNA (70-100 ng/μl) was co-injected with Texas Red lineage marker (Molecular Probes) into the yolk of 256-512 cell embryos.
LR scoring and whole-mount in situ hybridization (WISH)
Zebrafish cardiac jogging and looping was assessed by light microscopy or by WISH with nkx2.5. Gut looping was determined by foxA3 at 48-52 hpf. Xenopus heart and gut orientation was scored under a light microscope (stage 44-46) and laterality was monitored by xnr1expression (stage 22-24). For all manipulations, embryos were fixed in 4%paraformaldehyde/1×PBS. Digoxygenin-labeled RNA probes (Roche) were synthesized using linearized templates and the appropriate RNA polymerase. Zebrafish and Xenopus hybridizations were done as described(Long et al., 2003; Houston and Wylie, 2005).
TopFlash assay
Zebrafish embryos were injected with either TopFlash or FopFlash as previously described (Park and Moon,2002), incubated until epiboly stage and treated with thapsigargin or DMSO for 10 minutes at 60% epiboly. After three to five washes with embryo medium, excess medium was removed and embryos were flash frozen at the indicated time (20 or 60 minutes after drug application). Triplicate samples of 15 embryos each were used for TopFlash/FopFlash dual luciferase assays performed following manufacturer instructions (Promega). Luminescence was detected in a Turner Biosystems 20/20n Luminometer.
Immunolocalization
Immunofluorescence was performed according to a standard protocol using anti-acetylated tubulin (Sigma), β-catenin (Sigma) or Ntl (a gift from Dr D. J. Grunwald, University of Utah) followed by fluorescent secondary antibody(Alexa Fluor 488, Molecular Probes) or conjugated anti-horseradish peroxidase(Jackson ImmunoResearch). Confocal image stacks collected at 2 μm intervals were evaluated for nuclear β-catenin.
Syto-11
Syto-11 (Molecular Probes) was diluted in embryo medium to a final concentration of 7.5 μM. Control solution was 2% DMSO in embryo medium. Zebrafish (70% epiboly) and Xenopus (stage 11.5) were soaked in Syto-11 for 30 minutes and washed out with excess medium.
Calcium imaging
Zebrafish embryos (one-cell) injected with Fura-2 (dextran-conjugated,Molecular Probes) were oriented with the dorsal shield location noted before collection. The instrumentation for data collection, image analysis software and data manipulations were as described(Slusarski and Corces,2000).
RESULTS
Localized DFC Ca2+ release
As the morphogenetic movements of epiboly progresses, additional cell movements of gastrulation occur, such as involution, convergence and extension, to produce the primary germ layers and the embryonic axis. Ca2+-release activity has been described during epiboly/gastrula stages in both zebrafish and Xenopus(Gilland et al., 1999; Wallingford et al., 2001; Webb and Miller, 2006). During epiboly, zebrafish embryos are staged by the extent to which cells have moved over the yolk. At 50% epiboly, the embryonic dorsal shield forms as a thickening on the future dorsal side(Kimmel et al., 1995). The DFCs lie just below the shield and migrate ahead of the shield during gastrulation. At the completion of gastrulation, the DFCs come to lie between the yolk and tailbud and will form the epithelial lining of the KV.
To visualize spatial and temporal changes in live embryos during epiboly/gastrulation, we used the ratiometric Ca2+ sensor Fura-2, a reliable measure of intracellular Ca2+(Grynkiewicz et al., 1985). As described in the introduction, Ca2+ indicators have been used to investigate the role of Ca2+ in LR patterning. Most of the commonly used Ca2+ indicators utilize single wavelength excitation and have the potential to interpret signal artifacts as Ca2+-dependent changes. Dual-wavelength detection of Fura-2 has reduced sensitivity to signal artifacts and enables quantitative measurement of Ca2+concentrations. In addition, Fura-2 is less prone to photobleaching (the irreversible destruction of fluorophores) allowing for extended imaging over a developmental time point. Using Fura-2 imaging in zebrafish, we detected localized Ca2+ release activity in the shield/DFC region between 60-90% epiboly. In a lateral view of a bright field image, the DFCs resided at the leading edge ahead of the shield region(Fig. 1A, arrow). A single frame of the raw Fura-2 fluorescence of an embryo in a lateral orientation with the shield to the right (Fig. 1B, arrow) and the associated ratio image(Fig. 1C, arrow) demonstrates localized Ca2+ release activity in the region of the DFCs. The transient Ca2+ release in the DFC region was most active during 60-70% epiboly and became less active after 75-80% epiboly (see Movie 1 in the supplementary material).
Ca2+ manipulation impacts organ laterality. (A-C)Lateral view, dorsal to the right of 70-80% epiboly stage embryos with arrows indicating the DFC region in (A) bright field image, (B) an individual frame of Fura-2 (380) fluorescence and (C) the corresponding ratiometric image(340/380). The ratio image is converted to pseudo-color with yellow indicating high free intracellular Ca2+. (D) Summary plot of thapsigargin exposure vs heart jogging defects. (E) Summary of heart jogging defects in wt, control and after 10 and 20 minutes of thapsigargin treatment at 60% epiboly. Lateral view of wt (F) and thapsigargin-treated (G) embryos. (H) Summary of organ laterality defects. Thap, thapsigargin.
Ca2+ manipulation impacts organ laterality. (A-C)Lateral view, dorsal to the right of 70-80% epiboly stage embryos with arrows indicating the DFC region in (A) bright field image, (B) an individual frame of Fura-2 (380) fluorescence and (C) the corresponding ratiometric image(340/380). The ratio image is converted to pseudo-color with yellow indicating high free intracellular Ca2+. (D) Summary plot of thapsigargin exposure vs heart jogging defects. (E) Summary of heart jogging defects in wt, control and after 10 and 20 minutes of thapsigargin treatment at 60% epiboly. Lateral view of wt (F) and thapsigargin-treated (G) embryos. (H) Summary of organ laterality defects. Thap, thapsigargin.
Next, we determined whether a specific developmental process was impacted by experimental suppression of Ca2+ release during this time window. To this end, we used thapsigargin, a membrane permeable Ca2+-ATPase inhibitor that prevents the pumping of Ca2+back into the endoplasmic reticulum, resulting in transient increases in cytoplasmic Ca2+, but ultimately causing depletion of internal Ca2+ stores (Thastrup et al.,1990; Thastrup et al.,1994). Epiboly stage zebrafish embryos were incubated in thapsigargin at different doses (data not shown) and for different lengths of time (Fig. 1D). The optimal dose was determined to be 2.5 μM thapsigargin for 10 minutes and treatment after 60% epiboly, coinciding with the stages of local Ca2+ fluxes observed in the DFC region (see Movie 1 in the supplementary material). Embryos treated at that stage showed normal axial (anterior-posterior and dorsal-ventral) patterning (Fig. 1F,G), but displayed altered LR asymmetry as evaluated by heart jogging (see below; Fig. 1E). Doubling the exposure time did not have any substantial impact upon LR patterning above the optimal treatment(Fig. 1E). Thus, we conclude that brief exposure to thapsigargin after the 60% epiboly stage targets local short-term (or high threshold) signaling, leaving global Ca2+release relatively unperturbed and thereby allowing normal axial patterning. We next explored the nature of the laterality defects and whether DFC differentiation or morphogenesis was affected.
Ca2+ inhibition alters organ and molecular LR asymmetry
Heart jogging is one of the first morphological indications of organ LR asymmetry in zebrafish. At 24-30 hpf, the heart tube elongates from the midline and bends, or `jogs', to the left. To evaluate heart asymmetry embryos can be scored for leftward, rightward or no jog bias(Chen et al., 1997). In untreated, wild-type (wt), embryos, the heart jogged to the left in 95% of the embryos (Fig. 1H and Fig. 2A). In thapsigargin-treated embryos, jogging was randomized, as the heart tube migrated leftward (43%), rightward (25%), or remained in the middle (32%)(Fig. 1H and Fig. 2A-C). A conserved aspect of LR asymmetry among vertebrates is a later morphological looping of the heart, which typically forms a D-loop that places the zebrafish ventricle anterior and to the right of the atrium. In thapsigargin-treated embryos there was concordance between cardiac jogging and looping (see Table S1 in the supplementary material).
If Ca2+ inhibition during epiboly disrupts a key LR signal, then the placement of other asymmetrically oriented organs, in addition to the heart, should be altered. This could manifest itself either as heterotaxia(organs adopting laterality independently), or as concordance, which can lead to situs inversus totalis (reversal of all organ LR asymmetries). Embryos sorted by heart jogging were stained for foxA3 expression, a larval gut marker (Fig. 1H and Fig. 2D-F) and we found a tight correlation between heart jogging and gut positioning. These results show that organ laterality defects tend to be coordinated in thapsigargin-treated embryos.
Thapsigargin treatment disrupts organ laterality and molecular asymmetry. (A-C) Dorsal view, 30 hpf, showing nkx2.5expression in the heart tube, denoting a left jog in wt (A), and no jog (B) or right jog (C) in thapsigargin-treated embryos. (D-F) Dorsal view, 48 hpf, white arrows indicate foxA3 gut expression in wt (D) and symmetrical (E) or reverse loop (F) in thapsigargin-treated embryos.(G-I) Dorsal view, 19-23 somites, arrows indicate left-sided spaw expression in (G) wt, and (H) bilateral or (I) right-sided expression in thapsigargin-treated. Dorsal view, 19-23 somites, arrows indicate LPM, dashed arrows indicate brain lefty1 expression on the left side in (J) wt, and (K) bilateral in thapsigargin-treated,arrowheads denote lefty1 expression in the midline. Lateral view of ntl expression in (L) wt and (M) thapsigargin-treated,and shh expression in (N) wt and (O)thapsigargin-treated embryos. (P) Summary of asymmetrical marker expression.
Thapsigargin treatment disrupts organ laterality and molecular asymmetry. (A-C) Dorsal view, 30 hpf, showing nkx2.5expression in the heart tube, denoting a left jog in wt (A), and no jog (B) or right jog (C) in thapsigargin-treated embryos. (D-F) Dorsal view, 48 hpf, white arrows indicate foxA3 gut expression in wt (D) and symmetrical (E) or reverse loop (F) in thapsigargin-treated embryos.(G-I) Dorsal view, 19-23 somites, arrows indicate left-sided spaw expression in (G) wt, and (H) bilateral or (I) right-sided expression in thapsigargin-treated. Dorsal view, 19-23 somites, arrows indicate LPM, dashed arrows indicate brain lefty1 expression on the left side in (J) wt, and (K) bilateral in thapsigargin-treated,arrowheads denote lefty1 expression in the midline. Lateral view of ntl expression in (L) wt and (M) thapsigargin-treated,and shh expression in (N) wt and (O)thapsigargin-treated embryos. (P) Summary of asymmetrical marker expression.
To determine whether the randomized organ asymmetry is preceded by alteration of underlying molecular cues, we analyzed several LR markers. The zebrafish nodal-related gene southpaw (spaw)(Long et al., 2003) is the earliest asymmetrical marker in the left LPM during somitogenesis (14-21 hpf). At the 4-6 somite stage, spaw expression initiated flanking the KV and this early expression was maintained in thapsigargin-treated embryos (see Fig. S1 in the supplementary material). However, the later asymmetric spaw expression in the LPM was dramatically altered in thapsigargin-treated embryos with only a small percentage showing left-sided expression (Fig. 2G-I) compared with wt and control embryos (Fig. 2P). We next investigated markers induced by asymmetric spaw, including lefty1, lefty2 and pitx2c (reviewed by Ahmad et al., 2004). In wt and control embryos, lefty1 was expressed in the left anterior LPM(overlapping the prospective heart field)(Fig. 2J), but in treated embryos, expression was predominantly bilateral(Fig. 2K) or reversed. In the developing brain, pitx2c and lefty1 were transiently expressed in the left dorsal diencephalon(Halpern et al., 2003). Wt and control embryos showed left-sided lefty1(Fig. 2J), whereas thapsigargin-treated embryos showed altered lefty1 brain expression(Fig. 2K). Likewise, lefty2 and pitx2c expression was severely disrupted in treated embryos (Fig. 2P and see Fig. S1 in the supplementary material). Thus, thapsigargin treatment disrupts normal asymmetric expression of known molecular cues.
The dorsal midline (notochord and floorplate) is important as a barrier for laterality signals and any gaps in the midline barrier can lead to bilateral spaw expression (Bisgrove et al.,2000; Danos and Yost,1996; Meno et al.,1998). Thus, we evaluated midline integrity in treated embryos. DIC optics in live embryos (Fig. 1G, data not shown) as well as prechordal/notochord expression in the goosecoid-GFP transgenic line(Doitsidou et al., 2002)showed normal midline patterning (see Fig. S2 in the supplementary material). At the molecular level, we evaluated three independent notochord and/or floorplate markers (lefty1, ntl and shh) and all three were normal in thapsigargin-treated embryos(Fig. 2J-O). In the case of ntl, we evaluated both RNA (Fig. 2L,M) and protein distribution (see Fig. S2 in the supplementary material). Since ntl, lefty1 and shh have essential functions associated with the midline and were expressed, these analyses confirm that brief (10 minute) thapsigargin treatment does not compromise the integrity of dorsal midline tissues.
Analysis of DFC Ca2+ source
Thapsigargin treatment of zebrafish embryos yielded dramatic alteration to laterality without alteration to the midline. Typically thapsigargin treatment results in inhibition of Ca2+ release. Nevertheless, other studies use thapsigargin to increase Ca2+ levels, in particular those that investigate capacitative Ca2+ release. To determine whether thapsigargin led to an increase or a loss of the DFC regional Ca2+activity, we performed Fura-2 image analysis in zebrafish embryos at the time of treatment. After application of thapsigargin, we observed suppression of the DFC-regional Ca2+ fluxes (see Movie 2 in the supplementary material). The loss of the DFC Ca2+ release activity upon thapsigargin treatment supports a role for intracellular Ca2+ in a subset of cells during epiboly.
The DFCs are specified and migrate to the tailbud in thapsigargin-treated embryos.sox17 expression at 80% epiboly in(A) wt and (B) thapsigargin-treated embryos. Enriched Ntl protein at 80% epiboly in the DFCs in (C) wt and (D)thapsigargin-treated embryos; arrowheads note the DFC region. Expression of sox17 at 4-5 somites in (E) wt and (F) XeC-treated embryos; asterisks indicate dispersed cells, arrows indicate DFC/KV region. Tg (Dusp6:d2EGFP) expression at 80% epiboly in (G) untreated and (H) thapsigargin-treated embryos; arrows indicate DFCs. EGFP expression at tailbud to 1-somite in (I) untreated and(J) thapsigargin-treated embryos and at 5 somites in (K)untreated and (L) thapsigargin-treated embryos. XeC, Xestospongin C; n,notochord.
The DFCs are specified and migrate to the tailbud in thapsigargin-treated embryos.sox17 expression at 80% epiboly in(A) wt and (B) thapsigargin-treated embryos. Enriched Ntl protein at 80% epiboly in the DFCs in (C) wt and (D)thapsigargin-treated embryos; arrowheads note the DFC region. Expression of sox17 at 4-5 somites in (E) wt and (F) XeC-treated embryos; asterisks indicate dispersed cells, arrows indicate DFC/KV region. Tg (Dusp6:d2EGFP) expression at 80% epiboly in (G) untreated and (H) thapsigargin-treated embryos; arrows indicate DFCs. EGFP expression at tailbud to 1-somite in (I) untreated and(J) thapsigargin-treated embryos and at 5 somites in (K)untreated and (L) thapsigargin-treated embryos. XeC, Xestospongin C; n,notochord.
To substantiate the necessity of DFC regional Ca2+, we tested additional reagents that suppress Ca2+ release including:cyclopiazonic acid, a reversible inhibitor of Ca2+-ATPase(Seidler et al., 1989), and valproate, an inhibitor of inositol turnover, targeting the phosphatidylinositol (PI) cycle (Eickholt et al., 2005) as well as Xestospongin C (XeC), a compound that inhibits IP3-induced Ca2+ release(Gafni et al., 1997). All three reagents, cyclopiazonic acid, valproate and XeC, alter heart jogging in treated embryos (data not shown). Embryos treated with valproate and evaluated for asymmetrical gene expression demonstrated altered leftyexpression (see Fig. S3 in the supplementary material). Taken together, these data support a requirement for Ca2+ release in the DFCs for appropriate laterality.
Ca2+ inhibition disrupts late aspects of DFC development
Since the DFCs are located in the region of endogenous Ca2+release activity, we investigated DFC specification, migration and endocytic activity after thapsigargin treatment. Several molecular markers are expressed in the DFCs during epiboly (squint, left-right dynein-related-1/lrdr1,ntl and sox17) (Alexander and Stainier, 1999; Essner et al.,2002; Feldman et al.,1998; Schulte-Merker et al.,1994). We focused on these markers because three of these genes(lrdr1, ntl and squint) are required for proper LR patterning and sox17 can be used to evaluate morphogenesis. In thapsigargin-treated embryos, robust DFC sox17 expression was observed at 80% epiboly in a linear domain as opposed to an ovoid shape observed in controls (Fig. 3A,B). lrdr1 expression at 80% epiboly was similar in wt and thapsigargin-treated embryos (see Fig. S4 in the supplementary material). squint expression, a zebrafish nodal-related gene, was also maintained (data not shown). DFCs expressed both ntl RNA (see Fig. S4 in the supplementary material) and Ntl protein in thapsigargin-treated embryos but the cells did not expand over the yolk as much as in the control(Fig. 3C,D). The presence of early DFC markers supports the fact that Ca2+ inhibition does not interfere with the specification of the DFCs. However the compact DFC domain at later epiboly suggests that DFC behavior and or migration may be altered.
Since DFCs are highly endocytic and readily take up vital dyes, such as Syto-11 (Cooper and D'Amico,1996), we determined whether this aspect of DFC function was affected. Control and thapsigargin-treated embryos incubated in Syto-11 showed similar DFC dye uptake, as visualized by fluorescent microscopy. In thapsigargin-treated embryos, Syto-11 uptake and early migration to the tailbud region was similar to control-treated embryos (see Fig. S4 in the supplementary material). We next monitored DFC migration in vivo by using a GFP transgenic line that marks the DFCs. In the Dusp6:d2EGFPtransgenic line, EGFP was expressed in a fibroblast growth factor(FGF)-responsive manner in several tissues including the KV and its progenitors (Molina et al.,2007). We directly visualized DFC migration in live embryos from 80% epiboly to four-somite stage (see Movie 3 in the supplementary material). Migration during epiboly was similar in wt and thapsigargin-treated embryos(Fig. 3G,H). By the tailbud to one-somite stage in wt, the DFCs formed a circular cluster at the midline(Fig. 3I) whereas thapsigargin-treated embryos displayed a smaller midline cluster with individual cells scattered nearby (Fig. 3J and see Movie 4 in the supplementary material). By the three-to five-somite stage in wt, the DFCs formed the circular precursor of the KV(Fig. 3K), but in thapsigargin-treated embryos, there was increased dispersal(Fig. 3L).
Evaluation of DFC markers, lrdr1 and sox17, demonstrated a similar migration pattern. DFC migration in thapsigargin-treated embryos appeared the same as in the control during early epiboly but showed increased dispersal of individual cells during somite stages (see Fig. S4 in the supplementary material). Use of another inhibitor, XeC, also resulted in increased dispersal of DFCs, as assayed by the distribution of sox17(Fig. 3F). sox17 is expressed in the endoderm as well as in the DFCs and endodermal expression is similar in control and thapsigargin-treated embryos(Fig. 3A,B). Thus, transient Ca2+ inhibition specifically affects the DFCs, and not overall sox17 expression. These data support the conclusion that inhibition of Ca2+ release does not perturb DFC specification, endocytic activity or early migration, but rather is required for later migration or coalescence.
Inhibition of DFC regional Ca2+ release suppresses KV formation
The DFCs are KV progenitors (Cooper and D'Amico, 1996; Melby et al.,1996), and thapsigargin treatment resulted in a dramatic loss of a visible KV, with 97% severely reduced or absent(Fig. 4A-D). Thapsigargin-treated embryos also displayed reduced or absent cilia in the KV region (see Fig. S4 in the supplementary material). A subset of KV cells expressed Charon, a Nodal antagonist of the Cer/Dan protein class(Hashimoto et al., 2004). During somite stages, charon expression appeared as a horseshoe shape in the tailbud (Fig. 4E)whereas the pattern in thapsigargin-treated embryos was severely reduced(Fig. 4F) or absent(Fig. 4A,G). Thus, loss of the endogenous Ca2+ flux during epiboly eliminates charonexpression and disrupts formation of the KV.
DFC regional Ca2+ is required for KV morphogenesis.(A) Summary of KV defects. (B-D) Bright field image of wt embryo with normal KV (B) and thapsigargin-treated with reduced (C) and absent (D)KV: dorsal view, 10 somites. Arrows indicate vesicle location; n, notochord.(E-G) charon expression (arrows) around KV in wt (E) and thapsigargin-treated embryos with reduced (F) and absent (G) signal: lateral view, 10 somites. Insets show dorsal view of charon expression. Thap,thapsigargin. Scale bars: 50 μm.
DFC regional Ca2+ is required for KV morphogenesis.(A) Summary of KV defects. (B-D) Bright field image of wt embryo with normal KV (B) and thapsigargin-treated with reduced (C) and absent (D)KV: dorsal view, 10 somites. Arrows indicate vesicle location; n, notochord.(E-G) charon expression (arrows) around KV in wt (E) and thapsigargin-treated embryos with reduced (F) and absent (G) signal: lateral view, 10 somites. Insets show dorsal view of charon expression. Thap,thapsigargin. Scale bars: 50 μm.
Ca2+ signaling negatively controls β-catenin activity
In zebrafish, we previously demonstrated an antagonistic relationship between Ca2+ signaling and β-catenin activation(Westfall et al., 2003b). Consequently, we explored the possibility that β-catenin stability may be a target of Ca2+ signaling in the DFC region. After 10 minutes of thapsigargin treatment, embryos were washed, cultured until 80% epiboly (50 minutes after treatment) and fixed for immunolocalization. Confocal analysis of β-catenin distribution on both sides of control-treated and wt embryos displayed sporadic cells with nuclear β-catenin in the enveloping layer(EVL) and the yolk syncytial layer (YSL)(Fig. 5A). The EVL, which forms an epithelial monolayer covering the blastoderm, and the YSL, which resides at the blastoderm/yolk margin, are regions where endogenous Ca2+release has previously been detected(Reinhard et al., 1995; Slusarski et al., 1997) and both layers contribute to the DFCs (Cooper and D'Amico, 1996). A marked increase of β-catenin-positive nuclei was observed in thapsigargin-treated embryos(Fig. 5B). Focusing on the DFC-region, there was little or no nuclear β-catenin in wt embryos(Fig. 5C). In contrast, the DFC region of thapsigargin-treated embryos displayed nuclear β-catenin(Fig. 5D). Student's t-test analysis of the average number of β-catenin-positive nuclei demonstrated a statistically significant, eightfold increase, in thapsigargin-treated compared with untreated embryos(Fig. 5F). To confirm that ectopic nuclear β-catenin is transcriptionally active, we used TopFlash reporter analyses. TopFlash or FopFlash control vectors were injected into embryos, which were then treated with thapsigargin or control reagent. At 80%epiboly, reporter activity was assayed via relative luminescence units(normalized for Renilla control) and TopFlash-injected embryos, but not FopFlash, showed a striking, statistically significant increase after thapsigargin treatment (Fig. 5G). The increased reporter activity at this stage supports the conclusion that thapsigargin treatment leads to transcriptionally active,ectopic β-catenin. Increased TopFlash activity was observed as early as 20 minutes after thapsigargin application (see Fig. S5 in the supplementary material), suggesting a direct response to Ca2+ inhibition. For in vivo confirmation of β-catenin activation, we used transgenic embryos expressing destabilized GFP under the control of a β-catenin-responsive promoter, TOPdGFP (Dorsky et al.,2002). In wt embryos at 80% epiboly, there was active Wnt/β-catenin signaling in ventral-lateral domains, reflected by robust TOPdGFP expression, and minimal signal in the DFC region. Thapsigargin treatment induced ectopic GFP expression in the DFCs and flanking regions (see Fig. S5 in the supplementary material).
To determine whether activated β-catenin is sufficient to disrupt LR patterning we targeted a stabilized β-catenin construct [βcat55-C(Pelegri and Maischein, 1998)]to the DFCs, using a procedure developed by Amack and Yost(Amack and Yost, 2004). In this method, reagents injected into the yolk of 256- to 512-cell stage embryos were preferentially taken up by the YSL and DFC progenitors, which maintain cytoplasmic bridges with the yolk cell (DFC-targeting). In our hands, the injections successfully targeted the DFCs with 20-40% efficiency as measured by control GFP RNA or co-injected fluorescent tracer and we assumed a similar efficiency with unlabelled reagents. DFC-directed βcat55-C RNA was sufficient to cause abnormal heart jogging and altered lefty1/2expression, whereas control RNA injection had no change in lefty1/2expression (Fig. 5H,I). To confirm that the induced defects were due to activated β-catenin at biologically relevant levels, we utilized DFC-targeted Axin1-MO(antisense morpholino oligo), a known negative regulator of β-catenin(Heisenberg et al., 2001). Axin1 knockdown in the DFC region leads to ectopic nuclearβ-catenin (Fig. 5E). However, the modest impact of Axin1 knockdown on lefty1/2expression reflects the efficiency of the DFC-targeting; this effect was significant and was sufficient to alter KV formation (see Fig. S5 in the supplementary material). Therefore activated β-catenin in the DFCs is sufficient to alter KV formation and subsequent organ laterality.
To verify that Ca2+ activity in the DFC region is critical for KV formation, we utilized DFC targeting of additional Ca2+ release inhibitors. KV formation was normal in wt or control-injected (DFC targeting of DMSO) embryos (see Fig. S5 in the supplementary material). Consistent with global thapsigargin treatment, DFC targeting of thapsigargin, or valproate(inositol turnover inhibitor) disrupted KV formation (see Fig. S5 in the supplementary material). Taken together, these data affirm that the defects we observe are primarily due to intracellular Ca2+ release inhibition in the DFC region.
DFC-like cells in Xenopus
Our data provide evidence that early Ca2+ release in the zebrafish DFC region is necessary for proper KV formation and subsequent LR asymmetry by antagonism of β-catenin. In Xenopus, several pieces of evidence suggest that early Ca2+ activity near cells with DFC-like properties could be a conserved mechanism for establishing laterality. First, Ca2+ release activity has been described near the late organizer region in Xenopus gastrula stage embryos(Webb and Miller, 2006). Second, stage 13 Xenopus embryos have a cluster of left-right dynein-expressing cells in the dorsal blastopore(Essner et al., 2002) and finally, stage 15 embryos have recently been shown to have motile cilia with left-ward fluid flow in a node-like structure(Schweickert et al., 2007). Thus, we explored the possibility that Xenopus embryos have a population of endocytically active cells, which ultimately contribute to the tailbud and form a ciliated structure, analogous to zebrafish DFCs(Cooper and D'Amico, 1996). Syto-11 uptake in late gastrula embryos (stage 11.5) revealed accumulation of labeled cells in the dorsal mesoderm of the tail region of embryos sagittally bisected at stage 14 (Fig. 6A)and stage 17 (neurula) (Fig. 6B). Archenteron roof tissue explants from stage 17 embryos showed Syto-11 staining of the node (Fig. 6C, arrow) and surrounding lateral plate mesoderm. Subsequent immunostaining of the explants for acetylated tubulin showed enrichment of cilia in the node region (Fig. 6D, arrow). These data establish that highly endocytic cells in the late gastrula preferentially contribute to the node region in Xenopus.
Ca2+ inhibition activates β-catenin andβ -catenin is sufficient to alter laterality. Confocal images ofβ-catenin immunolocalization at 70-80% epiboly in (A) wt and(B) thapsigargin-treated embryos. HRP staining of β-catenin protein oriented on the DFC region in (C) wt, (D)thapsigargin-treated and (E) DFC-targeted Axin-1 MO-injected embryos;arrowheads indicate β-catenin-positive nuclei. (F) Comparison of total nuclear β-catenin in wt and thapsigargin-treated embryos.* indicates P-value of 2.5×10-5.(G) TopFlash (top) vs. FopFlash (fop) luciferase reporter constructs analyzed for relative luminescence in wt or thapsigargin-treated embryos,normalized to control (Renilla) luciferase. Data are means ±s.e. (H,I) Combined lefty1 and lefty2expression (arrows) in the left LPM and brain in wt (H) and bilateral (I) in DFC-targeted β-catenin RNA-injected embryos: dorsal-anterior view, 19-23 somites.
Ca2+ inhibition activates β-catenin andβ -catenin is sufficient to alter laterality. Confocal images ofβ-catenin immunolocalization at 70-80% epiboly in (A) wt and(B) thapsigargin-treated embryos. HRP staining of β-catenin protein oriented on the DFC region in (C) wt, (D)thapsigargin-treated and (E) DFC-targeted Axin-1 MO-injected embryos;arrowheads indicate β-catenin-positive nuclei. (F) Comparison of total nuclear β-catenin in wt and thapsigargin-treated embryos.* indicates P-value of 2.5×10-5.(G) TopFlash (top) vs. FopFlash (fop) luciferase reporter constructs analyzed for relative luminescence in wt or thapsigargin-treated embryos,normalized to control (Renilla) luciferase. Data are means ±s.e. (H,I) Combined lefty1 and lefty2expression (arrows) in the left LPM and brain in wt (H) and bilateral (I) in DFC-targeted β-catenin RNA-injected embryos: dorsal-anterior view, 19-23 somites.
Ca2+ inhibition induces ectopic nuclear β-catenin and laterality defects in Xenopus
To determine the extent to which these endocytically active cells share Ca2+ sensitivity, we manipulated Ca2+ release in Xenopus, at a comparable stage of gastrulation (stage 11.5), and using a similar limited exposure as in zebrafish. We visually scored heart and gut laterality after stage 45 (Yost,1992). Normal organ placement (situs solitus), with the heart ventricle situated on the left side and the outflow tract looping to the right and concomitant counterclockwise coiling of the gut, was observed in 100% of wt and control-treated embryos (Fig. 6E, arrow and M). In thapsigargin-treated embryos, we observed heart-looping and gut-coiling defects (Fig. 6F, arrow and M). Endogenous Xenopus nodal-related 1(xnr1) expression is detected asymmetrically in the left LPM at stages 22-24 (Fig. 6G,M). Thapsigargin-treated embryos displayed abnormal xnr1 expression(Fig. 6H,M), showing loss of xnr1 or bilateral expression in the posterior LPM. Furthermore,thapsigargin treatment was sufficient to induce ectopic nuclear β-catenin on both sides of a stage 13 Xenopus embryo compared with the control(Fig. 6I-J and see Fig. S6 in the supplementary material) and as in zebrafish, brief thapsigargin treatment did not alter overall axial anterior-posterior and dorsal-ventral patterning(Fig. 6K,L). In conclusion,these data show that Xenopus embryos have endocytically active cells at the late gastrula stage, which contribute to tissues in the tail, similar to zebrafish DFCs. Also, a short pulse of Ca2+ inhibition during Xenopus gastrulation is sufficient to generate ectopic nuclearβ-catenin before node formation and cause laterality defects as observed in zebrafish. These results suggest a conserved Ca2+-sensitive event involving cells with DFC-like properties in LR patterning in non-mammalian vertebrates.
DISCUSSION
Here we report endogenous local Ca2+ release in the zebrafish DFC region and show that brief inhibition of Ca2+ release during epiboly results in altered DFC coalescence/cohesion, defective KV formation,and disrupted expression of asymmetrical LR signals as well as concordant alteration of organ laterality. We identify β-catenin activity as a candidate target of this Ca2+ signaling. These observations link endogenous local Ca2+ fluxes and β-catenin regulation to the proper establishment of the KV and subsequent LR asymmetry in the zebrafish embryo. Additionally, we define a DFC-like cell population in Xenopusembryos and demonstrate that these cells share a similar Ca2+-sensitive stage, and exhibit ectopic nuclear β-catenin resulting in altered laterality upon Ca2+ perturbation. Our results in both zebrafish and Xenopus identify an early Ca2+-sensitive phase relevant to LR patterning that precedes the Ca2+ fluxes linked to cilia-driven nodal flow.
Xenopus DFC-like cells and Ca2+ requirement for laterality.Xenopus embryos incubated with Syto-11 at stage 11.5 and sagittally bisected at (A) stage 14 and (B) stage 17 showing Syto-11-labeled cells enriched in the future tail region. (C) Stage 17 archenteron roof explant, showing Syto-11 staining of the node (arrow) and surrounding LPM. (D) Explant in C immunostained for acetylated tubulin(acTub), which is enriched in the node region (arrow). (E) Ventral view of stage 46 control-treated embryo, showing situs solitus, or normal lateral asymmetry. The heart is outlined, showing the left-sided position of the ventricle and right-sided outflow tract. The gut coils counterclockwise,as shown by the arrow. (F) Ventral view of a thapsigargin-treated embryo at stage 46, showing an example of situs inversus, or reversed laterality. The heart is outlined, showing the right-sided position of the ventricle and left-sided outflow tract. The gut coils clockwise, as shown by the arrow. (G) Left-sided xnr1 expression at stage 22-24 in a DMSO-treated embryo. (H) Bilateral xnr1 expression in a thapsigargin-treated embryo. (I,J) Lateral view ofβ-catenin protein localization in control-treated (I) and thapsigargin-treated (J) stage 13 embryos. Note increased number and intensity of nuclear staining. (K,L) Lateral views of control (K) and thapsigargin-treated (L) embryos, showing normal axial development. (M)Summary of Xenopus laterality data. Thap, thapsigargin.
Xenopus DFC-like cells and Ca2+ requirement for laterality.Xenopus embryos incubated with Syto-11 at stage 11.5 and sagittally bisected at (A) stage 14 and (B) stage 17 showing Syto-11-labeled cells enriched in the future tail region. (C) Stage 17 archenteron roof explant, showing Syto-11 staining of the node (arrow) and surrounding LPM. (D) Explant in C immunostained for acetylated tubulin(acTub), which is enriched in the node region (arrow). (E) Ventral view of stage 46 control-treated embryo, showing situs solitus, or normal lateral asymmetry. The heart is outlined, showing the left-sided position of the ventricle and right-sided outflow tract. The gut coils counterclockwise,as shown by the arrow. (F) Ventral view of a thapsigargin-treated embryo at stage 46, showing an example of situs inversus, or reversed laterality. The heart is outlined, showing the right-sided position of the ventricle and left-sided outflow tract. The gut coils clockwise, as shown by the arrow. (G) Left-sided xnr1 expression at stage 22-24 in a DMSO-treated embryo. (H) Bilateral xnr1 expression in a thapsigargin-treated embryo. (I,J) Lateral view ofβ-catenin protein localization in control-treated (I) and thapsigargin-treated (J) stage 13 embryos. Note increased number and intensity of nuclear staining. (K,L) Lateral views of control (K) and thapsigargin-treated (L) embryos, showing normal axial development. (M)Summary of Xenopus laterality data. Thap, thapsigargin.
Pan-embryonic Ca2+ waves during epiboly/gastrula stages are believed to coordinate convergent extension (CE), the lengthening and narrowing of groups of cells, as well as their directed migration to the future dorsal side (Wallingford et al.,2001; Wallingford et al.,2002). Genetic data supports this idea, in that zebrafish homozygous zygotic Wnt-5 mutants (pipetail) display CE defects and have reduced Ca2+ release frequency(Westfall et al., 2003a). In fact, thapsigargin treatment during 30-50% epiboly resulted in cell movement defects and a shortened anterior-posterior axis (data not shown), consistent with previously described thapsigargin-induced phenotypes(Creton, 2004). However, brief thapsigargin treatment of epiboly/gastrula stage zebrafish and Xenopus disrupts laterality without perturbing normal cell movement. Similar laterality defects are generated with additional inhibitors(valproate, to disrupt inositol levels, XeC, to inhibit IP3-induced Ca2+ release and cyclopiazonic acid, to deplete endoplasmic Ca2+ stores) and confirm the requirement for early intracellular Ca2+ release in KV formation. DFC-targeting of inhibitors supports the conclusion that the Ca2+ requirement resides in a small population of cells, as opposed to global signaling over the entire embryo.
Asymmetrical Ca2+ levels across the mouse and chick node have been implicated in LR axis determination(McGrath et al., 2003; Raya et al., 2004). The zebrafish KV contains monociliated cells similar to those found in the mouse node and several studies support a link between the node/KV and LR patterning(Brueckner, 2001; Essner et al., 2005; Hashimoto et al., 2004; Supp et al., 1999; Supp et al., 1997). Studies in zebrafish have established a role for inositol polyphosphates, important intracellular second messengers, in LR patterning. Inositol polyphosphate kinases (Ipk) use IP3 as a template to generate higher inositols(IP4, IP5 and IP6), and play key roles in maintaining Ca2+ homeostasis by regulating the concentrations of IP3 and other inositols (Xia and Yang, 2005). ipk1 knockdown in zebrafish disrupts laterality, resulting in right-sided or bilateral spaw expression(Sarmah et al., 2005). However, ipk1 is not expressed in the DFCs and in ipk1knockdown embryos, KV morphology is normal and cilia are present(Sarmah et al., 2005). Recent work has also postulated a distinct Ca2+ requirement in the regulation of KV cilia motility in zebrafish(Shu et al., 2007). These previous studies describe changes in laterality after KV formation. Our findings indicate a novel early role for Ca2+ in the DFCs for KV formation. Interestingly in our current study, we found that embryos treated with thapsigargin during tailbud and early somite stages did not have any impact upon KV formation or organ laterality. We used a fluorescent-labeled thapsigargin and noted that upon treatment during epiboly stages, labeled thapsigargin penetrated the YSL and DFC regions (data not shown). Conversely,treatment after tailbud and during somite stages revealed that labeled thapsigargin does not penetrate the region of the embryo where the KV resides. Thus a negative result with thapsigargin treatment at later stages (i.e. somite stages) may be a reflection of failure of the reagent to permeate into the target tissues.
We propose that epiboly stage Ca2+ release is required to regulate β-catenin activity in the DFC region. Ca2+ inhibition for a discrete time period resulted in a statistically significant increase in nuclear β-catenin and activation of β-catenin transcriptional reporters, including TOPdGFP transgenic embryos and TopFlash assays. In fact,we observe TopFlash activation as early as 20 minutes after addition of thapsigargin, and this activation occurs hours before somitogenesis, KV and cilia formation. This activation is confirmed in the whole embryo by use of the TOPdGFP transgenic line. We also show that DFC endocytic activity and marker gene expression (ntl, lrdr, sox17 and squint) are maintained in Ca2+-inhibited embryos. Based on in vivo imaging and Syto-11 staining, DFC cell number remained unchanged; however migration within the tailbud region was altered. The labeled cells in the tailbud appear dispersed and did not coalesce into a KV. This may be due to altered cell adhesion, polarized cell movement or failure to undergo a mesenchymal-to-epithelial transition. All are processes that can be perturbed with altered Wnt signaling or β-catenin levels.
Our data provide evidence for an early Ca2+-dependent cascade that ultimately regulates β-catenin levels and promotes KV formation. Murine Wnt3a has been shown to be necessary in LR coordination, but after the node cilia have formed (Nakaya et al.,2005). Our findings indicate a critical role for Ca2+release and β-catenin activity, specifically in the DFC region before somitogenesis. This may be mediated by Wnt/Ca2+ antagonism of Wnt/β-catenin (Slusarski et al.,1997; Torres et al.,1996), however the Wnt/Ca2+ ligand is not identified in this study. Since the DFCs do not form a KV in both thapsigargin-treated and Axin1-MO injected embryos, we cannot evaluate effects on cilia-dependent laterality cues, but we do observe a loss of charon expression. It should be noted that in charon knockdown embryos, the KV remains intact while laterality is altered (M.R., unpublished observations). Conversely, knockdown of Bardet-Biedl Syndrome genes in zebrafish embryos eliminates the morphological aspects of the KV and KV cilia but charon is still present (Yen et al., 2006). Furthermore, charon knockdown generates a preponderance of embryos with bilateral expression of laterality markers(Hashimoto et al., 2004). Since Ca2+ inhibition eliminates charon expression, this loss of charon can explain the high incidence of bilateral marker gene expression in thapsigargin-treated embryos in the presence of an intact midline.
The dramatic impact upon LR signaling caused by a transient pulse of Ca2+ inhibition would suggest a key patterning event and would also be predicted to be shared with other vertebrates. In Xenopus, we identified a population of endocytically active cells that ultimately surrounded the putative frog node, thus defining DFC-like properties in another vertebrate. Ca2+-release inhibitors targeting the equivalent developmental stage in Xenopus, induced ectopicβ-catenin, organ laterality defects and bilateral nodal(Xnr-1) expression.
In summary, we have identified an early local Ca2+ release in the DFC region and show that inhibition of this activity via DFC-targeted injections results in dramatic KV defects. Furthermore, we implicateβ-catenin activity as the target of the endogenous DFC Ca2+fluxes and we show that suppression of β-catenin signaling within the DFCs is necessary for appropriate LR patterning in zebrafish and Xenopus. Thus, antagonisms between Wnt/Ca2+ and Wnt/β-catenin pathways may be crucial not only for dorsal-ventral patterning but for KV morphogenesis and LR asymmetry.
Acknowledgements
We thank H. Griesbach and T. Westfall for technical assistance and Slusarski lab members for comments on the manuscript. This work was supported by the American Heart Association pre-doctoral fellowship to I.S., Carver Charitable Trust Grants to M.R.R. and D.W.H. and by National Institutes of Health grant CA112369 to D.C.S.