Cardiomyocytes (CMs) are nonregenerative. Self-renewable pluripotent human embryonic stem cells (hESCs) can differentiate into CMs for cell-based therapies. In adult CMs, Ca2+-induced Ca2+ release from the sarcoplasmic reticulum (SR) via the ryanodine receptor (RyR) is key in excitation-contraction coupling. Therefore, proper Ca2+ handling properties of hESC-derived CMs are required for their successful functional integration with the recipient heart. Here, we performed a comprehensive analysis of CMs differentiated from the H1 (H1-CMs) and HES2 (HES2-CMs) hESC lines and human fetal (F) and adult (A) left ventricular (LV) CMs. Upon electrical stimulation, all of H1-, HES2-, and FLV-CMs generated similar Ca2+ transients. Caffeine induced Ca2+ release in 65% of FLV-CMs and ∼38% of H1- and HES2-CMs. Ryanodine significantly reduced the electrically evoked Ca2+ transient amplitudes of caffeine-responsive but not -insensitive HES2- and H1-CMs and slowed their upstroke; thapsigargin, which inhibits the sarco/endoplasmic reticulum Ca2+-ATPase (SERCA) pump, reduced the amplitude of only caffeine-responsive HES2- and H1-CMs and slowed the decay. SERCA2a expression was highest in ALV-CMs but comparable among H1-, HES2-, and FLV-CMs. The Na+-Ca2+ exchanger was substantially expressed in both HES2- and H1-CMs relative to FLV- and ALV-CMs. RyR was expressed in HES2-, H1-, and FLV-CMs, but the organized pattern for ALV-CMs was not observed. The regulatory proteins junctin, triadin, and calsequestrin were expressed in ALV-CMs but not HES2- and H1-CMs. We conclude that functional SRs are indeed expressed in hESC-CMs, albeit immaturely. Our results may lead to driven maturation of Ca2+ handling properties of hESC-CMs for enhanced contractile functions.
Disclosure of potential conflicts of interest is found at the end of this article.
The primary function of our heart is to mechanically pump blood throughout the body. However, cardiomyocytes (CMs) are nonregenerative. As a result, transplantation is the last resort for end-stage heart failure patients, but this is hampered by the severe shortage of donor organs [1, 2]. Human (h) embryonic stem cells (ESCs), derived from the inner cell mass of human blastocysts, can self-renew while maintaining their pluripotency . Upon in vitro induction, hESCs can differentiate into spontaneously beating CMs [4, , , –8]. Indeed, hESC-derived CMs (hESC-CMs) display structural and functional properties of early-stage cardiomyocytes  and can functionally integrate with [4, 9] or even electrically pace the recipient heart after transplantation in vivo .
During an action potential of adult CMs, Ca2+ entry into the cytosol through sarcolemmal L-type Ca2+ channels triggers the release of Ca2+ from the intracellular Ca2+ stores (also known as sarcoplasmic reticulum [SR]) via the ryanodine receptor (RyR). This process, the so-called Ca2+-induced Ca2+ release (CICR) , escalates the cytosolic Ca2+ ([Ca2+]i) to activate the contractile apparatus for contraction. For relaxation, elevated [Ca2+]i is pumped back into the SR by the sarco/endoplasmic reticulum Ca2+-ATPase (SERCA) and extruded by the Na+-Ca2+ exchanger (NCX) to return to the resting [Ca2+]i level. Such a rise and subsequent decay of [Ca2+]i is known as Ca2+ transient . Given the central importance of CICR in cardiac excitation-contraction coupling, proper Ca2+ handling properties of hESC-CMs are therefore crucial for their successful functional integration with the recipient heart after transplantation. Indeed, abnormal Ca2+ handling, as in the case of heart failure, can even be arrhythmogenic (e.g., delayed after depolarization) [10, 12].
In mouse (m) ESC-CMs, both the SR load and RyR are essential for regulating contractions even at very early developmental stages . By contrast, it has been reported that spontaneously beating hESC-CMs (derived from the H9.2 and I3 hESC lines) do not have functional SRs and that their contractions result from trans-sarcolemmal Ca2+ influx rather than Ca2+ release from the SR . To better define the poorly known Ca2+ handling properties of hESC-CMs, here we performed a comprehensive analysis of Ca2+ transients recorded from CMs differentiated from the H1 (H1-CMs) and HES2 (HES2-CMs) hESC lines and compared their properties to those of human fetal left ventricular CMs (FLV-CMs, 16–18 weeks) under different electrophysiological and pharmacological conditions. Human FLV-CMs, which have been suggested as a choice for myocardial repair , were chosen for comparison because they are developing yet functional CMs. Furthermore, the electrical phenotypes of hESC-CMs have been reported to exhibit fetal-like properties [5, 6], but a detailed comparison of their Ca2+ handling properties has not been performed. Here, we provide experimental evidence that both H1- and HES2-CMs do express functional SRs. A better understanding of these fundamental properties of hESC-CMs is crucial for designing effective strategies or protocols for improving both their safety and functional efficacy (e.g., facilitated or driven maturation of Ca2+ handling properties for enhanced contractile functions).
Materials and Methods
hESC Culturing and Differentiation
The HES2 (ESI, Singapore, http://www.escellinternational.com) and H1 (WiCell Research Institute, Madison, WI, http://www.wicell.org) hESC lines (NIH codes are ES02 and WA01, respectively) chosen for this study were cultured and differentiated as we previously described [4, 16, 17]. Briefly, HES2 cells were grown on mitomycin C (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) inactivated mouse embryonic feeders (mEFs). Culture medium consisted of Dulbecco's modified Eagle's medium (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) containing 2 mM l-glutamine, insulin-transferrin-selenium, nonessential amino acids, 90 μM β-mercaptoethanol, and 20% fetal bovine serum (HyClone, Logan, UT, http://www.hyclone.com). HES2 cells were passaged manually (“cut and paste”) by cutting colonies into pieces and removing them from the mEFs using dispase (8 mg/ml; Invitrogen). For cardiac differentiation, HES2 cells were removed from the mEFs, resuspended, and broken into pieces, followed by coculturing with the immortalized endoderm-like END2 cells at 100% confluence .
H1 cells were grown on irradiated mEFs from 13.5-day embryos of CF-1 mice and propagated as previously described . The culture medium consisted of 80% knockout Dulbecco's modified Eagle's medium, 20% knockout serum replacement, 4 ng/ml basic fibroblast growth factor, 1 mmol/l glutamine, 0.1 mmol/l β-mercaptoethanol, and 1% nonessential amino acid solution (all from Gibco-BRL, Gaithersburg, MD, http://www.invitrogen.com). To induce the formation of EBs, hESCs were detached using 1 mg/ml type IV collagenase (Gibco-BRL) and transferred to Petri dishes containing 80% knockout Dulbecco's modified Eagle's medium, 20% fetal bovine serum defined, 1 mmol/l glutamine, and 1% nonessential amino acid stock in the absence of β-fibroblast growth factor. The aggregates were cultured in suspension for 7 days, followed by plating on gelatin-coated (0.1%; Sigma-Aldrich) 6-well plates to form hESC-CMs.
Isolation of hESC-CMs
For isolating hESC-CMs, beating outgrowths were microsurgically dissected from H1- (7 + 11 to 17 or 18–24 days) and HES2- (18–24 days) derived EBs by a glass knife , followed by incubation in collagenase II (1 mg/ml) at 37°C for 30 minutes. The isolated cells were incubated with KB solution containing (mM): 85 KCl, 30 K2HPO4, 5 MgSO4, 1 EGTA, 2 Na2-ATP, 5 pyruvic acid, 5 creatine, 20 taurine, and 20 d-glucose at room temperature for 30 minutes. After the cells were plated on laminin-coated glass coverslips for 1 hour at 37°C, regular culture medium was added. Same as previous reports [6, 18], sarcomeres were displayed in these derived CMs as shown after myosin heavy chain, α-actinin, and tropomyosin staining, although much less evident and defined than those in adult CMs. Calcium recordings from cell clusters containing 10–15 cells were performed within 48 hours after plating.
Isolation of Human Fetal and Adult Left Ventricular Cardiomyocytes
Human FLV-CMs and adult-CMs were isolated and experimented on according to protocols approved by the UC Davis Institutional Review Board (protocol numbers 200614787-1 and 200614594-1). Briefly, fetal human hearts (16–18 weeks; Advanced Bioscience Resources, Alameda, CA, http://www.abr-inc.com) and adult human hearts (18+ years; National Disease Research Interchange, Philadelphia, http://www.ndriresource.org) were perfused with enzymatic solutions using a customized Langendorff apparatus as previously described . FLV-CMs were cultured on laminin-coated glass coverslips in 24-well dishes with a density of ∼5 × 105 cells per well in a water-jacket incubator at 37°C with medium containing 5 mM carnitine, 5 mM creatine, 5 mM taurine, 0.5% penicillin-streptomycin (Gibco-BRL) and 10% fetal bovine serum in Medium 199 (Sigma-Aldrich). Adult-CMs were stored at −80°C for Western blotting.
Measurements of Cytosolic Ca2+
A spectrofluorometric method with Fura-2/AM as the Ca2+ indicator was used for measuring [Ca2+]i. FLV- or hESC-CMs were incubated with 5 μM Fura-2/AM and 0.02% Pluronic F-127 for 30 minutes at 37°C. Fluorescent signals obtained upon excitation at 340 nm (F340) and 380 nm (F380) were recorded from cells perfused with Tyrode solution containing (mM): 140 NaCl, 5.0 KCl, 1.0 CaCl2, 1.0 MaCl2, 10.0 glucose, and 10 HEPES (pH 7.4) unless otherwise indicated. Data were analyzed using the IonWizard software (version 5; IonOptix, Milton, MA, http://www.ionoptix.com) to generate the Ca2+ transient parameters reported in this study. The F340/F380 ratio was used to represent cytosolic [Ca2+]i. To induce cytoplasmic Ca2+ transients, CMs were stimulated by electrically pulsing from 0.1–0.5 Hz or by caffeine application as indicated. For electrical stimulations, Ca2+ transients were recorded and analyzed after a series of depolarizations that enabled each transient to fully decay so as to establish a steady-state SR content.
Cells were fixed for 15 minutes at room temperature with 4% paraformaldehyde in phosphate-buffered saline (PBS). After washing with PBS, cells were permeabilized in PBS containing 0.2% Triton X-100. Primary mouse anti-RyR monoclonal antibody (MA3-925; Affinity BioReagents, Golden, CO, http://www.bioreagents.com) was diluted with 10% goat serum at 1:100. Alexa Fluor 488 anti-mouse IgG (A-11029; Invitrogen) was the second antibody used for fluorescence imaging. Hoechst 33342 (H3570; Invitrogen) was used to stain the nuclei. Coverslips were mounted onto glass slides in Prolong Gold antifade reagent (Invitrogen). Samples were imaged on a confocal laser-scanning microscope (C1si; Nikon, Tokyo, http://www.nikon.com).
Proteins (12 μg) were loaded in SDS-polyacrylamide (10%) gel and separated by electrophoresis at 150 V for 2 hours. The separated proteins were transferred electrophoretically from the gel onto nitrocellulose membrane at 100 V at 4°C for 1 hour in a buffer containing 25 mM Tris-base, 192 mM glycine, and 20% methanol. After the membranes were washed in a buffer (Tris-buffered saline [TBS], pH 7.4, containing 0.1% Tween 20 and 5% skimmed milk powder) for 60 minutes at room temperature to block nonspecific binding, they were probed at 4°C overnight with anti-SERCA2a (ab2861; Abcam, Cambridge, MA, http://www.abcam.com), anti-NCX1 (ab2869; Abcam), anti-calsequestrin (ab3516; Abcam), anti-triadin (sc-33391; Santa Cruz Biotechnology, Santa Cruz, CA, http://www.scbt.com), anti-junctin (sc-33367), or anti-calreticulin (ab22683; Abcam), respectively. After washing for 30 minutes with TBS (0.1% Tween 20 solution), the membranes were then incubated for 1 hour with a secondary antibody solution conjugated to horseradish peroxidase-conjugated rabbit anti-mouse at 1:2,000 dilution. Then the membranes were washed for 30 minutes with TBS. Detection was performed with an ECL Plus Western blotting detection system (GE Healthcare, Giles, UK, www.gehealthcare.com).
All data were expressed as means ± SEM. One-way analysis of variance followed by Newman-Keuls multiple comparison tests or paired t test was carried out to test for differences between the mean values within the same study. A difference of p < .05 was considered significant.
Electrically Evoked Ca2+ Transients of hESC- and FLV-CMs Had Similar Properties
Figure 1A and 1B shows that the basal cytosolic Ca2+, an index of Ca2+ homeostasis regulated by various Ca2+ handling proteins (such as the RyR, SERCA2a, etc.), was significantly lower in HES2-CMs (n = 17) than in FLV-CMs (n = 15). However, no detectable significant difference was observed between H1- (n = 18) and FLV-CMs (p > .05). Upon electrical stimulation, all of HES2-, H1-, and FLV-CMs examined similarly generated Ca2+ transients with statistically identical amplitude, maximum upstroke velocity (Vmax, upstroke), and maximum decay velocity (Vmax, decay) (p > .05; Fig. 1C–1E). The experiments that follow were designed to further explore the basis of and any latent differences in the Ca2+ handling properties of HES2-, H1-, and FLV-CMs.
Differential Responses of hESC- and FLV-CMs to Caffeine
To investigate whether functional SRs are indeed expressed in HES2-, H1-, and FLV-CMs and their Ca2+ contents, we next studied the effect of caffeine, an opener of RyR, on cytosolic Ca2+. To exclude the contribution of trans-sarcolemmal Ca2+ influx via voltage-gated Ca2+ channels that have been shown to express in hESC-CMs , the experiments were performed in the absence of Ca2+ in the extracellular bath. Figure 2A and 2B shows that a brief exposure to caffeine (10 mM) induced a rise in cytosolic Ca2+ that subsequently decayed back to the baseline in 65% of FLV-CMs (n = 11 of 17). By contrast, only 35% (n = 7 of 20) and 40% (n = 8 of 20), respectively, of H1- and HES2-CMs that generated Ca2+ transients upon electrical stimulation (Fig. 1) also elicited caffeine-induced Ca2+ transients. Despite the lower percentages of caffeine-responsive HES2- and H1-CMs relative to FLV-CMs, the caffeine-induced Ca2+ transient amplitudes were not different among themselves (p > .05; Fig. 2C). Thus, caffeine-responsive hESC-CMs had developed SR loads similar to that of FLV-CMs. Kinetically, H1-CMs displayed the highest Vmax, upstroke (Fig. 2D), but those of HES2- and FLV-CMs were comparable. As for the decay, FLV-CMs were most rapid followed by H1- then HES2-CMs (Fig. 2E). These functional differences are further explored below.
Effects of Ryanodine and Thapsigargin on Ca2+ Transients of hESC-CMs
Our caffeine experiments presented above clearly demonstrate that SRs in HES2- and H1-CMs were indeed expressed and operable. To relate the SR function of HES2- and H1-CMs to Ca2+ handling proteins such as RyR and SERCA2a, we next examined the effects of their specific inhibitors ryanodine  and thapsigargin , respectively, on electrically evoked Ca2+ transients. Figure 3A and 3B shows that, after application of 10 μM ryanodine for 30 minutes, the electrically evoked Ca2+ transient amplitudes of caffeine-responsive HES2- and H1-CMs were significantly reduced by 37% ± 4.8% and 18% ± 4.3%, respectively (p < .05; Fig. 3B, open bars). However, the amplitudes of caffeine-insensitive HES2- and H1-CMs were not affected by ryanodine (Fig. 3B, solid bars). Ryanodine also significantly slowed the Vmax, upstroke of caffeine-responsive but not caffeine-insensitive cells (Fig. 3C). Taken collectively, the above observations were consistent with the notion that functional RyRs were present only in caffeine-responsive cells.
In adult human CMs, SERCA2a is responsible for ∼70% of Ca2+ uptake from the cytoplasm back into the SR . Figure 4A and 4B shows that thapsigargin application (0.5 μM, 15 minutes) significantly reduced the electrically evoked Ca2+ transient amplitude of caffeine-responsive HES2- and H1-CMs. This was probably due to inhibited SR Ca2+ reload as a result of SERCA2a blockade by thapsigargin. In accordance with this notion, Vmax, decay of both HES2- and H1-CMs was significantly slowed by thapsigargin (Fig. 4C).
Ca2+ Handling Proteins in hESC-, FLV-, and Adult Left Ventricular-CMs
Figure 5A shows a representative Western blot analysis of SERCA2a and NCX in HES2-, H1-, and FLV- as well as human adult left ventricular- (ALV) CMs. All of the H1-, HES2-, and FLV-CMs expressed comparably high levels of SERCA2a, consistent with the responses of their Ca2+ transients to thapsigargin. As anticipated, the expression level of SERCA2a was highest in ALV-CMs [22, 23]. Unlike SERCA2a, NCX displayed a different protein expression profile. NCX was most abundant in FLV-CMs but only very weakly expressed in ALV-CMs, consistent with previously published results . Interestingly, NCX was substantially expressed in both HES2- and H1-CMs relative to ALV-CMs but much less so in comparison with FLV-CMs.
As for RyR, immunostaining was performed. Figure 5B shows that RyR was indeed expressed in HES2-, H1-, and FLV-CMs. However, the organized, regularly spaced expression pattern as previously reported for adult human ventricular cardiomyocytes  was not observed. In the junctional SR membrane, RyR forms a macrocomplex with several regulatory proteins including junctin (Jn), triadin (Trd), and calsequestrin (CSQ). Figure 5C shows that all of Jn, Trd, and CSQ were expressed in ALV-CMs but not HES2- and H1-CMs. As for FLV-CMs, CSQ and Trd but not Jn were expressed but still at levels substantially less than those of ALV-CMs. Developmentally, immature CMs are known to express significant levels of calreticulin; calreticulin decreases after birth due to post-transcriptional modification and is subsequently replaced by CSQ during SR maturation [26, –28]. As anticipated from these previous results, Figure 5D shows that calreticulin was abundantly and comparably expressed in all of HES2-, H1-, and FLV-CMs. Taken collectively, our results indicate that SR-related proteins in human heart cells undergo substantial developmental changes.
Characterizing the functional properties of hESC-CMs is a crucial first step for their eventual clinical applications for myocardial repair. Although recent studies have revealed several important cellular electrical properties of hESC-CMs [5, 6, 29], their Ca2+ handling properties are much less defined, and the availability of relevant data is extremely scarce with only one published report to date  (the reported data have been reviewed [30, 31]). As mentioned, proper Ca2+ handling is crucial for the successful functional integration of hESC-derived cardiac grafts after transplantation and for ensuring their lack of arrhythmogenicity. In brief, the major findings of our present study are as follows. First, in contrast to the previous report , our new data support that functional SRs (i.e., RyR and SERCA2a) are indeed expressed in hESC-CMs; thus, CICR contributes to Ca2+ transients even at early developmental stages, like the murine ESC-CMs . Second, human ESC-CMs that evoke electrically induced Ca2+ transients consist of caffeine-responsive and -insensitive cells (with and without functional SR, respectively), probably due to the presence of differentiating CMs of different developmental stages. Third, SERCA2a is expressed in hESC-CMs but at a level substantially less than the adult counterpart; by contrast, NCX is expressed at a higher level in hESC-CMs than ALV-CMs. Fourth, the SR-associated Ca2+ handling regulatory proteins triadin, calsequestrin, and junctin are expressed in ALV- but not hESC-CMs. These findings are further discussed below (a) in comparison to previously published results so as to provide a better basic understanding of the Ca2+ handling properties of hESC-CMs and (b) in relation to the development of novel strategies to facilitate the maturation of hESC-CMs for improving their functional efficacy for therapies.
Same as the study by Dolnikov and colleagues , Ca2+ transients could be readily generated from both HES2- and H1-CMs upon electrical stimulations. Unlike the previous report, however, we found that at least two subpopulations, caffeine-responsive and -insensitive, were present in hESC-CMs and FLV-CMs. Caffeine induces large Ca2+ transients in ∼38% of hESC-CMs, indicating that this caffeine-responsive subpopulation expresses functional SRs and RyRs that are capable of loading and unloading Ca2+. The percentage of caffeine-responsive cells is higher in FLV-CMs (∼65%). The difference could be attributed to the presence of a larger population of developmentally immature hESC-CMs with un- or underdeveloped SR; indeed, increased SR load has been suggested to improve the efficacy of voltage-gated Ca2+ currents as a trigger for SR Ca2+ release for effective excitation-contraction coupling [32, –34]. Although ∼85% and 60% of HES2- and H1-CMs, respectively, belong to the ventricular type, atrial and pacemaker derivatives are also known to be present in spontaneously contracting human embryoid bodies [5, 6, 16]; this heterogeneity of chamber-specific cells likely further contributes to the lower percentages of caffeine-sensitive hESC-CMs. Human ESC lines whose cardiac derivatives have been genetically labeled, such as that recently described by Huber et al. , will be useful tools for distinguishing among these possibilities.
Dolnikov et al. reports that neither ryanodine nor caffeine affects Ca2+ transients of hESCs . This apparent difference is indeed consistent with our data and can be readily accounted for by the caffeine-responsive population newly identified in the present study. In our recordings, only Ca2+ transients of caffeine-responsive but not -insensitive cells can be functionally inhibited by ryanodine. Immunostaining confirms the expression of RyR proteins in hESC-CMs. However, the expression pattern is distinct from the highly organized distribution seen in adult cardiomyocytes  but similar to that of FLV-CMs. Although human and mouse (m) ESCs (and their cardiac derivatives) differ in many important ways, the developmental aspect of the Ca2+ handling properties of hESCs resembles that of mESCs: RyRs are expressed in very early stages and can be caffeine-induced to lead to Ca2+ transients for contractions . Of note, 18- to 24-day-old hESC-CMs were investigated in the present study. According to Sartiani et al. , these hESC-CMs can be considered as early CMs (15–40 days) whose electrophysiological properties are relatively immature. This notion is consistent with our observation that only ∼38% of hESC-CMs expressed functional SR. Nonetheless, sarcomeres were displayed as shown by myosin heavy chain, α-actinin, and tropomyosin staining, although much less evident and defined than those in adult CMs. Furthermore, although our cells were chronologically younger than those investigated by Dolnikov and colleagues (55-day-old H9.2-CMs) , we observed relatively more mature Ca2+ handling properties (as gauged by their responsiveness to caffeine). Collectively, the differences between their study and ours could be attributed to the different culturing, differentiation, and experimental conditions (e.g., clusters of 10–15 cells rather than the entire beating outgrowths from intact hEBs were chosen for our experiments) as well as other intrinsic differences between the different hESC lines studied (H9.2 and I3 vs. H1 and HES2 of our experiments). Further investigations of the basis of such phenotypic differences might lead to novel strategies for driven maturation.
In mESC-derived CMs, it has been suggested that spontaneous Ca2+ transients are triggered by inositol-1,4,5-trisphosphate (IP3)-mediated Ca2+ release, which is then amplified and modulated by RyR-mediated Ca2+ release . IP3 receptor is highly expressed in conductive CMs in either embryonic or adult hearts . Considering the role of IP3 in automaticity and generation of arrhythmias, IP3-sensitive stores may play an important role in hESC-CMs, but further experiments will be required to test this notion.
Immature Ca2+ handling properties of hESC-CMs can cause poor functional integration with the host myocardium at best or lethal arrhythmias at worst. Thus, it is desirable to develop methods for facilitating their maturation ex vivo. Since RyR and SERCA2a are already expressed, targeted expression of the regulatory proteins that are largely absent in hESC-CMs (such as junctin, triadin, calsequestrin, and phospholamban) via gene transfer or protein transfection might more effectively render their SR and Ca2+ handling properties mature or adult-like. Additionally, NCX is highly expressed in hESC- and FLV-CMs but not in ALV-CMs. It has been reported that NCX expression in human heart developmentally peaks at 20-week gestation and is substantially higher than that in adult heart ; the reduction of NCX expression may be a compensatory response to the increased SERCA activity. Thus, our in vitro experiments with hESC-CMs were also consistent with these previous results. The possibility of suppressing NCX activity in hESC-CMs in order to achieve the high SERCA2a:NCX ratio in adult CMs for driven maturation and for maintaining calcium homeostasis requires further investigation.
We conclude that functional SRs are present in hESC-CMs, albeit immaturely. Driven maturation may be achieved by targeted expression of specific Ca2+ handling proteins.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
This work was supported by Grants from the NIH (R01 HL72857 to R.A.L.), the Stem Cell Program of the University of California, Davis School of Medicine (to R.A.L.), and the California Institute for Regenerative Medicine, UC Davis Stem Cell Training Program Fellowship (to J.D.F.). C.W.S. was supported by a postdoctoral fellowship award from the Croucher Foundation.