Zebrafish runx1 promoter-EGFP transgenics mark discrete sites of definitive blood progenitors

Blood cells in the early vertebrate embryo arise from successive waves of hematopoiesis.^1,2  Within each wave, distinct lineages of hematopoietic cells differing in differentiation potential are generated in embryonic sites that are divergent among vertebrate species.³  In zebrafish, as in mammals, immediate emergence of hematopoietic cells from mesoderm constitutes primitive hematopoiesis. The mammalian extraembryonic yolk sac gives rise to the earliest macrophages and erythrocytes.⁴  Zebrafish primitive macrophages and erythrocytes derived from the lateral plate mesoderm (LPM; equivalent to mammalian yolk sac), specifically from the cephalic mesoderm and intermediate cell mass (ICM), respectively.^5,6  In contrast, a feature of definitive hematopoiesis is cell- lineage generation from multipotent stem cells or progenitors.¹  In recent years, the aorta-gonad-mesonephros (AGM) region has been thought to be the origin of persisting definitive hematopoietic stem cells (HSCs), with expression of genes required for HSC development in that environment.^7,8  In zebrafish, the AGM equivalent is located in the region of the ventral wall of the dorsal aorta (DA), where HSCs are identified by expression of genes, such as runx1 and c-myb.^9,10  Embryonic HSC differentiation to produce multiple lineages proceeds when cells from the AGM migrate to other hematopoietic niches.¹¹  Zebrafish AGM progenitors first migrate to the posterior ICM, known as the posterior blood island (PBI) or caudal hematopoietic tissue (possible equivalent to mammalian fetal liver), before seeding the pronephros (mammalian bone marrow equivalent) and thymus.^12,13  The idea of having exclusive sites for primitive and definitive hematopoiesis has been challenged by several mammalian studies demonstrating a contribution of yolk-sac progenitors to the formation of definitive erythroid and myeloid lineages, most probably before HSCs arise.^4,14-16  This notion was corroborated in zebrafish by the discovery of erythromyeloid progenitors (EMPs) that arise within the PBI derivative of the LPM before emergence of HSCs in the wall of the DA.¹⁷  Furthermore, recent studies in mutant mice with either a defect in circulation or the migration of hematopoietic progenitors showed that normal hematopoiesis proceeded in the yolk sac, but definitive hematopoietic progenitors were absent in the AGM and fetal liver, raising the possibility of a yolk-sac origin for all definitive blood cells.^18,19  Our understanding of where HSCs arise therefore continues to be challenged, and there is much to learn about how the different hematopoietic microenvironments participate in promoting the adult blood program

The transcription factor Runx1 is regarded as a key regulator of definitive hematopoiesis, with a conserved gene structure and function in all vertebrates.^20,21  Runx1 is expressed in all sites of hematopoietic development in mammals and zebrafish, and is expressed in HSCs and the endothelial cells from which HSCs are thought to arise.^10,22,23  Homozygous inactivation of Runx1 in mice results in embryonic lethality, with a complete absence of definitive hematopoiesis.²²  Runx1 also functions in the maturation of megakaryocyte and T-cell lineages.²⁴  The 2 main isoforms of Runx1, generated by alternative usage of the promoters P1 and P2, differ in the 5′ untranslated region (UTR) and the coding sequence of the first exon.^25,26  This alternative promoter usage regulates Runx1 function in several ways. In addition to the differential temporal and spatial pattern of transcription,²⁷  some of the splice variants generated from P2 lack the transactivation domain and are thought to have dominant-negative functions in inhibiting myeloid differentiation.²⁶  The short 5′UTR in transcripts from the P1 promoter controls cap-dependent translation, whereas the long 5′UTR in P2 transcripts regulates internal ribosomal entry site–dependent translation.²⁸ 

With the aim of further advancing our understanding of the role of Runx1 in HSC biology, we generated 2 stable transgenic zebrafish lines expressing enhanced green fluorescent protein (EGFP) driven by either the distal P1 or proximal P2 zebrafish runx1 promoters. We show that segregated and successive emergence of definitive blood progenitors is regulated, in part, by alternative use of runx1 promoters. At 18 hours postfertilization (hpf), the Tg(runx1P1:EGFP)–transgenic line first marks EMPs that populate the PBI. By 22 hpf, the Tg(runx1P2:EGFP)–transgenic line displays EGFP expression within HSCs as they emerge from the zebrafish AGM. This line also later highlights blood cells within the head kidney and thymus.

Zebrafish (Danio rerio) embryos were obtained from natural spawnings and raised in embryo medium (E3) at 28.5°C. All the studies conducted for this manuscript have been reviewed and approved by the University of Auckland Animal Ethics Committee. A 12-kb fragment containing runx1P1 was identified with the DIG system for nonradioactive Southern blotting (Roche Diagnostics, Indianapolis, IN). The runx1P1 fragment was cloned into pBluescript II KS(+) (pBSIIKS+; Stratagene, La Jolla, CA). The 1.5-kb fragment at the 3′ end of the promoter sequence containing the runx1 coding sequence was removed by partial digest with SpeI. To position the runx1P1 promoter in-frame with EGFP, the 1.5-kb sequence adjacent to the coding sequence was polymerase chain reaction (PCR)–amplified with the primers EcoRIrunx1P1F and BamHIrunx1P1R, digested with EcoRI and BamHI, and then cloned into pT2KXIGΔin.²⁹  This was linearized with XhoI (blunted) and EcoRI, then ligated with the remaining 10.5-kb runx1P1 fragment that had been cut with SacII (blunted) and EcoRI, generating the runx1P1::EGFP/pT2K plasmid that was used for transgenesis. An 8-kb runx1P2 fragment was isolated by restriction digest with SacI and BamHI and cloned into the SacI/BamHI sites of pBSIISK+. The EGFP gene from pIRES-EGFP (Clontech, Mountain View, CA) was released by BamHI/XhoI digestion and cloned into the BamHI/XhoI sites of runx1P2/pBSIISK+. The constructs were injected into embryos at the one-cell stage and screened for germline integration by random intercrosses. The transgenic founders identified were crossed to establish stable transgenic lines.

runx1 morpholinos (MOs): runx1P1splice MO ATCGCTAGCAAACAACTCACCCCTG, runx1P1 ATG MO GGAGCGTCTGCTTGTCCACGTCTTC, and runx1 runt domain MO TTTTCAGCATCTCACCTCGTCCGCT (Gene Tools, Philomath, OR) were injected at a dose of 0.25 to 1.0 pmol per embryo.³⁰  MOs targeting gata1 and scl were used as described.^31,32 

Juvenile and adult fish were fixed in 4% paraformaldehyde, decalcified in 0.5 M ethylenediaminetetraacetic acid, dehydrated in ethanol, cleared in xylol, and then infiltrated and embedded in paraffin; 5-μm sagittal sections were cut and used for immunohistochemistry for EGFP expression as described,³³  and in situ hybridization for runx1. For detection of runx1 mRNA, sections were deparaffinized and rehydrated in phosphate-buffered saline. Sections were refixed in 4% paraformaldehyde, treated with 0.2 M HCl, 10 mg/mL proteinase K, followed by 0.25% acetic anhydride in 0.1 M triethanolamine, pH 8. Sections were hybridized overnight with a runx1 DIG riboprobe at 65°C.¹⁰  Unbound probes were washed with 2× saline sodium citrate and 0.2× saline sodium citrate. Sections were blocked, then incubated with anti-DIG–alkaline phosphatase. After washes, the sections were stained with nitro blue tetrazolium/5-bromo-4-chloro-3-indoyl phosphate p-toluidine, counterstained with Vector Nuclear Fast Red (Vector Laboratories, Burlingame, CA), and mounted in Entellan (Merck, Whitehouse Station, NJ).

Live transgenic embryos were imaged under a Leica MZ16FA fluorescence stereomicroscope with a DC490 camera (Leica Microsystems, Heerbrugg, Switzerland). Histologic sections were imaged with a Leica DMR compound microscope and a DC200 camera. For confocal imaging, embryos were anesthetized in tricaine, mounted in 1% (wt/vol) low-melt agarose in E3, and imaged with an Olympus FV1000 confocal microscope (Tokyo, Japan).

Transcriptional control of Runx1 from the 2 alternative promoters is thought to be important for modulating Runx1 function. Previous studies of Runx1 isoform expression in cell lines and mouse tissues suggested that HSCs and lymphocytes predominantly express the P1 isoform.^25,27  However, in zebrafish we have previously shown by MO knockdown studies that Runx1P2 is required for definitive hematopoiesis.¹⁰  Support for a role for the P2 isoform in early hematopoiesis and T-cell development has also recently come from mouse studies using a hypomorphic Runx1 allele with diminished P2 activity.³⁴  In Figure 1A, we show, by reverse-transcription PCR (RT-PCR), different temporal transcription profiles for the 2 isoforms during zebrafish embryogenesis. Only runx1P1 is maternally expressed, whereas significant levels of runx1P2 transcript are detected from 14 hpf.

To generate zebrafish with marked HSCs and to further investigate whether P1 and P2 have distinct regulatory functions in blood development, we cloned the 2 zebrafish runx1 promoters and generated stable transgenic lines expressing EGFP (Figure 1B). Bacterial artificial chromosomes (BACs) that included the runx1 locus were identified by Southern blotting to the BAC Zebrafish High-Density Filter Library (Incyte Genomics, Wilmington, DE). Four BACs were found to contain runx1 sequence 14i20, 74j23, 97a02, and 135g16. Although the full mRNA sequence of the runx1P2 isoform had been cloned, the fragment encoding the N-terminal region of the P1 isoform had not been identified when we commenced the study reported here. We used 5′ rapid amplification of cDNA ends to clone the sequence of runx1 unique to the P1 isoform. To amplify P1 isoform-specific mRNA sequence, degenerate primers were designed based on the highly conserved N-terminal sequence of mammalian RUNX genes. This resulted in a 93-bp sequence that was used to BLAST the Ensembl zebrafish genome (http://www.ensembl.org/Danio_rerio/index.html) to obtain further genomic sequences of the runx1P1 promoter. Approximately 1700 bp of the sequence upstream of the translational start site was available. Southern blot analysis was used to identify and clone further upstream promoter and enhancer sequences. A 73-bp sequence just upstream of the ATG was used as a probe to hybridize with runx1P1 sequences in the 97a02 BAC, which had been digested with NheI 69 bp downstream of the ATG. The 12-kb fragment identified as containing runx1P1 sequences was cloned (Figure 1B). The 3′ 1.5-kb sequence of the fragment was replaced with a PCR-amplified sequence, without the coding region, and cloned into the Tol2 vector.³⁵  The Tol2 runx1P1::EGFP vector was coinjected with transposase mRNA into one cell–stage embryos. The injected embryos were grown to adulthood and screened for germline integration by random intercrosses. Nineteen pairs of fish were successfully crossed, with 15 crosses producing transgenic embryos. A total of 16 transgenic founders were identified from further outcrosses. These transgenic lines displayed fluorescent expression in the same spatial domains.

We have previously shown, by morpholino knockdown studies, that the Runx1 isoform expressed from the P2 promoter is required for definitive hematopoiesis,¹⁰  suggesting that this promoter may be useful in generating a transgenic zebrafish line with labeled definitive blood. Restriction mapping of the 14i20 BAC was used to identify an 8-kb fragment containing runx1P2 promoter and enhancer sequences. This fragment was subsequently cloned into pBSIISK+ vector with EGFP and used to generate stable transgenic lines. The injected fish were screened by random intercrosses, and 5 transgenic founders were identified from the 92 fish screened. The different lines express EGFP in the same spatial and temporal domains but with different intensity.

Of the 2 transgenic lines, Tg(runx1P1:EGFP) is the first to express EGFP in hematopoietic tissues. Fluorescent cells were visible in the posterior LPM at 18 hpf (Figure 2A,B). Posterior LPM expression of runx1 is observed at 14 hpf¹⁰ ; the delay in EGFP detection was attributed to protein maturation. We detected EGFP transcripts in the LPM at 16 hpf (Figure S1, available on the Blood website; see the Supplemental Materials link at the top of the online article). EGFP expression was transient and by 24 hpf, when the PBI has formed from the posterior LPM, only weak expression remained (Figure 2C). The ontogeny of these cells resembles that of the definitive EMPs, which express erythroid and myeloid genes such as gata1 and pu.1, respectively.¹⁷  To confirm that the P1 promoter drives expression in EMPs, compound Tg(runx1P1:EGFP)/Tg(gata1:DsRED)³⁶  zebrafish were generated. At 24 hpf, cells in the PBI expressing DsRED also expressed EGFP (Figure 2D). EGFP expression in this domain was not detected at 30 hpf (data not shown), suggesting that runx1P1 is expressed at an early stage of EMP development but is soon down-regulated.

To examine the expression of runx1 isoforms in the EGFP-positive cells of runx1P1-transgenic embryos, the caudal region of 22 hpf embryos was dissected and dissociated (Figure S2). The EGFP-expressing cells were isolated by fluorescence-activated cell sorting (FACS). RT-PCR on RNA extracted from the sorted cells showed the runx1P1 isoform is predominantly expressed. The weak expression of the runx1P2 isoform possibly arises from EGFP expression in neuronal and somitic tissues.

To determine the function of Runx1P1, transcripts were selectively ablated using either a splice or an ATG MO, followed by expression analysis of a panel of erythromyeloid markers at 18 hpf (primitive hematopoiesis) and 36 hpf (EMPs). Effective inhibition of splicing by the runx1P1 MO was assessed by RT-PCR (Figure 3A). Knockdown of Runx1P1 did not interfere with the expression of early hematopoietic genes, such as gata1, hhex, and scl in the ICM at 18 hpf (Figure 3B–D), showing that Runx1P1 is not involved in primitive hematopoiesis. The initiation of definitive hematopoiesis, marked by the expression of runx1 and c-myb, is also unaffected by Runx1P1 depletion (Figure 3E,F). Expression of erythroid and myeloid markers within the PBI at 30 and 36 hpf was also unaltered by ablation of Runx1P1 (Figure 3G–I). This suggests that, whereas runx1P1 is expressed in the LPM and PBI preceding the emergence of EMPs, it performs a role other than the specification and maintenance of EMPs. Using an MO that targets the splicing of an exon within the runt domain common to both runx1 isoforms, the development of primitive blood and EMPs was similarly unaffected (Figure 3J–Q). Other tissues that express EGFP in the runx1P1 transgenic are the olfactory placode and the hypochord dorsal to the PBI, as well as weak expression in neuronal and somitic tissues (Figure 2A,C).

In the Tg(runx1P2:EGFP) line, EGFP was first observed at the 6-somite stage in the notochord (Figure 4A). By the 12-somite stage, distinct neuronal expression was observed (Figure 4B). Hematopoietic EGFP expression is first observed in the ventral wall of the DA or AGM at 22 hpf, recapitulating the runx1 in situ expression pattern (Figure 4C). This expression before the establishment of circulation suggests that the EGFP-positive cells are generated from the ventral wall of the DA, rather than being seeded from another hematopoietic site in the embryo. EGFP was abundant in the brain and spinal cord (Figure 4C–F). No EGFP was detected in the PBI region at 24 hpf where runx1 mRNA is expressed from 18 to 24 hpf (Figure 4F), suggesting the P2 isoform is not expressed in EMPs.

A highly conserved intronic enhancer element that drives Runx1 expression only in hematopoietic domains in mouse embryos has recently been identified.³⁷  We performed a similar analysis for zebrafish and Fugu runx1 but found no homologous regions (Figure S3), suggesting that this enhancer may not be present in teleosts.

By 36 hpf, round cells marked by EGFP were visualized between the DA and posterior cardinal vein (PCV) in the AGM, and the caudal artery (CA) and caudal vein (CV) in the PBI (Figure 4G–L). Crossing the Tg(runx1P2:EGFP)–transgenic with the Tg(gata1:DsRED) line,³⁶  to highlight red fluorescent erythrocytes within vessels, confirmed the localization of EGFP-positive cells in the AGM (Figure 5A,B). These features resemble the hematopoietic progenitors identified by Murayama et al¹²  and labeled in the Tg(scl:GFP) line.³⁸  As in the Tg(scl:GFP) line, we also detected EGFP expression in endothelial cells of the trunk vessels by confocal microscopy (Figure 4J). Labeling of these endothelial cells may represent expression at levels undetectable by in situ hybridization or may reflect exclusion of a repressor element required to silence vascular expression. EGFP-positive cells from the trunk and tail of 48 hpf runx1P2 transgenic embryos were isolated by FACS (Figure S2). RT-PCR showed the EGFP-expressing cells expressed only the runx1P2 isoform.

To confirm that the fluorescent cells in the AGM were hematopoietic (Figure 5A,B), we used MOs targeting several early hematopoietic transcription factors. We used the runx1 runt domain MO, and deletion of this important exon was confirmed by RT-PCR (data not shown). Injection of the runx1 runt domain MO resulted in the almost complete loss of hematopoietic EGFP expression at 48 hpf (Figure 5D,E), but neuronal expression was retained (Figure 5D). The knockdown of Scl affects both primitive and definitive blood, as well as development of the DA, from which the runx1-expressing HSCs are thought to emerge.^32,39  As expected, injection of an scl MO resulted in loss of the hematopoietic EGFP-positive cells (Figure 5F,G), without affecting the neuronal domains (Figure 5F). Gata1 is required for directing differentiation toward the erythroid lineage at the expense of myeloid specification.^31,33,40  Gata1 knockdown has been shown to increase both the level and duration of runx1 expression in the ICM, but its effect on definitive hematopoiesis has not been studied. In Tg(runx1P2:EGFP) embryos injected with a gata1 MO, EGFP in the AGM at 48 hpf was increased compared with controls (Figure 5H,I). It appears that, as in the ICM, Gata1 negatively regulates the maintenance of definitive progenitors driven toward the myeloid lineage.

The hematopoietic expression in the runx1P1 transgenic precedes that of runx1P2 and appears to mark 2 distinct populations of definitive hematopoietic cells. To show that these 2 lineages arise independently, the development of cells in the PBI of 20 hpf runx1P1 embryos was traced with a photoactivatable rhodamine dextran (Figure S4). Cells in the PBI gave rise to circulating erythrocytes and the caudal vasculature but did not migrate to the AGM.

The migration of HSCs from the PBI has been traced to the thymus and pronephros at 5 days postfertilization (dpf) using photoactivatable fluorescein-dextran.^12,41  We have never detected runx1 expression by in situ hybridization in the thymus or the pronephros at this stage, possibly because runx1 expression is reduced to levels undetectable by this technique during HSC differentiation. EGFP stability, with half-life more than 24 hours, might enable blood progenitors to be visualized after migration to the thymus and pronephros. By confocal microscopy, EGFP-positive cells were present in the thymus at 4 dpf and in the pronephros at 5 dpf (Figure 6A,B).

Beyond the embryonic and larval stages, we observed continued EGFP expression in the pronephros and mesonephros of 1-month-old runx1P2 transgenics (Figure 6C,D), with the same domain marked by runx1 in situ hybridization in wild-type fish sections (Figure 6E–G). In runx1P2 adult transgenics, staining with anti-EGFP demonstrated expression of EGFP in a small subset of cells within the pronephros (Figure 6H–J). This recapitulates the runx1 in situ hybridization profile on tissue sections (Figure 6K–M). Similar analysis of transgene expression within the runx1P1 reporter line showed no hematopoietic expression in the adult (data not shown).

In the development of the hematopoietic system, different blood lineages develop in overlapping temporal and spatial domains. Determining their origin, potential, and relationship to other lineages has been a challenge. Several recent studies on the lineage relationships between the different hematopoietic populations have produced differing results. In a recent study of zebrafish–definitive hematopoiesis, a population of definitive EMPs were identified in the PBI that arise from the posterior LPM, separately from HSCs from the AGM region.¹⁷  Previously, it was thought that the PBI only contained definitive HSCs that migrate from the AGM to undergo further maturation, before migration to the thymus and pronephros.^12,41,42  It was also suggested that HSCs in the AGM traced at different time points have different differentiation potential,¹²  although this was not observed in another study.⁴¹ 

The runx1 promoter–transgenic lines reported in this study mark 2 distinct populations of definitive blood, strongly supporting the notion that distinct lineages of definitive hematopoietic progenitors arise independently in the zebrafish embryo. Together, the 2 lines recapitulate the endogenous expression of runx1 in hematopoietic tissues.¹⁰  The 2 lineages of definitive blood in zebrafish appear to be analogous to those of mammalian systems. Definitive progenitors from the yolk sac can differentiate into myeloid and erythroid, but not lymphoid, lineages in cocultures, whereas progenitors from the AGM of mouse embryos of the same age have lymphoid potential.⁴³ 

The Tg(runx1P1:EGFP) line labels EMPs in the PBI, whereas the Tg(runx1P2:EGFP) line marks definitive progenitors arising from the AGM. In the runx1P1 line, the onset of EGFP expression by in situ hybridization and by fluorescence in the posterior LPM matches expression of runx1 mRNA in hematopoietic domains at a similar time point.¹⁰  Other transgenic zebrafish lines have been generated that express fluorescent markers in definitive blood cells, including scl, lmo2, fli1, and gata1.^36,38,44,45  As all of these transcription factors are expressed in both primitive and definitive blood, fluorescent expression is present along the whole length of the LPM. This is in contrast to the posteriorly restricted expression in the runx1P1 line, showing expression specific to EMPs and not primitive erythrocytes. In addition, lmo2 and fli1 are strongly expressed in the developing vasculature.^44,45 

EMPs are thought to be present in the PBI between 24 and 48 hpf and arise from lmo2-expressing cells in the posterior LPM at 13 to 15 somites (16 hpf),¹⁷  the stage at which we first detect EGFP mRNA in the runx1P1 transgenic. This observation, together with the coexpression of gata1 in this region, suggests that EMPs express the Runx1P1 isoform from an early stage of development in the posterior LPM. However, Runx1P1 does not appear to be required for the specification or maintenance of EMPs. EMPs are present in the PBI transiently and are later replaced by HSCs derived from the AGM. These 2 waves of definitive hematopoiesis may differ in their dependence on Runx1. The Runx1P1 isoform does not appear to be expressed in definitive HSCs emerging from the AGM region, as no fluorescence was detected in the ventral wall of the DA and runx1P1 is not expressed at 48 hpf by RT-PCR, a stage when strong expression of runx1P2 is detected in HSCs in the AGM.

In the Tg(runx1P2:EGFP) line, putative HSCs in the ventral wall of the DA express EGFP from 22 hpf, corresponding to the stage at which definitive HSCs are thought to first be present in this region. By 36 hpf, fluorescent cells resembling those in fli1, scl, and CD41 promoter transgenics are visible in the AGM between the DA and PCV. Although it has been suggested that the PBI contains exclusively EMPs before 40 hpf,¹⁷  cells expressing runx1P2:EGFP are found in the AGM along the entire length of the embryo, including the area posterior to the yolk extension designated the PBI. This raises the possibility that a heterogeneous population of definitive blood cells with different developmental potential resides in overlapping spatial domains. This is consistent with the observation that hematopoietic cells expressing the fli1:EGFP transgene can be traced from 30 hpf as they migrate to the thymus from either the anterior or posterior region of the AGM.⁴¹ 

It has been shown that fluorescent hematopoietic cells in the AGM, expressing either fli1 or CD41, migrate first to the PBI/caudal hematopoietic tissue region and then to the thymus, consistent with an HSC identity.^41,42  This is supported by results here, where the development of fluorescently labeled cells in the AGM in the runx1P2 transgenic requires the function of transcription factors known to be required for HSC development. Fluorescent cells are found later in the thymus and pronephros, suggesting that in the runx1P2 transgenic, hematopoietic progenitors are fluorescently marked from the very early stage of development, with persistence of expression as they migrate to other hematopoietic organs for multilineage differentiation.

Several recent zebrafish studies have contributed significant insights into definitive hematopoiesis,^12,17  but understanding the ontogeny of the hematopoietic system is far from complete. These runx1-transgenic lines are the first ones characterized that delineate separate waves of definitive hematopoiesis in zebrafish and will be useful tools to further explore HSC biology.

The authors thank L. Khan for technical assistance, A. Mahagaonkar for managing the zebrafish facility, S. Edgar for FACS analysis support, and the University of Auckland Biomedical Imaging Research Unit for their expert support with confocal microscopy. The Tol2/transposase expression vectors and the Tg(gata1:DsRed) transgenics were gifts from K. Kawakami and D. Traver, respectively.

This work was supported by grants from the Auckland Medical Research Foundation and the New Economy Research Fund, Foundation for Research, Science and Technology. E.Y.N.L. is a Bright Futures Top Achiever Doctoral Scholar.

Contribution: E.Y.N.L. and M.V.F. designed and performed the experimental work, analyzed the data, and wrote the paper; J.Y.M.C., M.L.K.-Z., T.M.F., R.S.M., and C.J.H. performed some experiments; and P.S.C. and K.E.C. planned the project, analyzed the data, and edited the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Maria Vega Flores, Department of Molecular Medicine & Pathology, School of Medical Sciences, University of Auckland, Private Bag 92019, Auckland, New Zealand; e-mail: m.flores@auckland.ac.nz.