If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. § Present address: Dept. of Pediatric Oncology, Dana-Farber Cancer Inst., Boston, MA 02115.
ADAR1 (adenosine deaminase acting on RNA-1) is widely expressed in mammals, but its biological role is unknown. We show here by gene targeting that ADAR1 selectively edits in vivo two of five closely spaced adenosines in the serotonin 5-hydroxytryptamine subtype 2C receptor pre-mRNA of nervous tissue; and hence, site-selective adenosine-to-inosine editing is indeed a function of ADAR1. Remarkably, homozygosity for two different null alleles of ADAR1 caused a consistent embryonic phenotype appearing early at embryonic day 11 and leading to death between embryonic days 11.5 and 12.5. This phenotype manifests a rapidly disintegrating liver structure, along with severe defects in definitive hematopoiesis, encompassing both erythroid and myeloid/granuloid progenitors as well as spleen colony-forming activity from the aorta-gonad-mesonephros region and fetal liver. Probably as a consequence of these developmental impairments, ADAR1-deficient embryonic stem cells failed to contribute to liver, bone marrow, spleen, thymus, and blood in adult chimeric mice. Thus, ADAR1 subserves critical steps in developing non-nervous tissue, which are likely to include transcript editing.
From Caenorhabditis elegans to humans, adenosine-to-inosine (A-to-I) editing of primary transcripts in the nucleus recodes exonic information and can lead to structural and functional changes in the encoded protein, thereby enlarging the protein space encoded by the transcriptome (
). In mammals, only few edits have been identified, all by serendipity and all in brain-expressed transcripts. Two sequence-related genes encode candidate enzymes for A-to-I editing, termed ADAR1 and ADAR2, each endowed with regions for binding double-stranded RNA (dsRNA)
The single instance in which an ADAR protein could be assigned to a particular edit concerns ADAR2 and the transcript for the glutamate receptor subunit GluR-B, which is edited to nearly 100% in a particular codon for a critical ion channel determinant (
). This is the only example where editing is necessary for survival, whereas for the other known edited transcripts, site-selective A-to-I conversion may serve to fine-tune the function of the encoded proteins (
The biological function of ADAR1 and its involvement in transcript editing are still unresolved and have become even enigmatic with a report that haploinsufficiency of ADAR1 leads to defects in the hematopoietic system and an embryonic lethal phenotype (
). These features might indicate that the biological functions of ADAR1 exceed transcript editing. However, as of today, no molecular process with ADAR1 participation has been delineated in mammals, nor has ADAR1 been convincingly shown to edit any of the known sites in brain-expressed transcripts, although two observations provide suggestive evidence for at least one transcript. In ADAR2-deficient mice, two sites in the serotonin 5-HT2C receptor transcript change little in editing status (
To gain better insight into the biological role of ADAR1, we generated mice with adar1 null alleles. These alleles, in combination with an adar2 null allele, permitted us to use a genetic approach in the mouse to assign different, closely spaced edits in 5-HT2C receptor transcripts unambiguously to either ADAR1 or ADAR2 activity, thus demonstrating that site-selective A-to-I editing is indeed a biological function of ADAR1.
Heterozygosity in mice for differently configured adar1 null alleles is compatible with survival and a normal appearance, in stark contrast to the embryonic lethality claimed for haploinsufficiency of ADAR1 (
). With homozygosity for our adar1 null alleles, however, embryos at mid-gestation developed a severe liver defect and died between E11.5 and E12.5. Remarkably, the defect involved cells of both the hematopoietic and hepatic lineages, as further suggested by analysis of mouse chimeras derived from ADAR1-deficient ES cells. Hence, ADAR1 has a critical role in non-nervous tissue, but edited transcripts remain to be identified.
adar1 Gene Targeting—The targeting vector for Cre/loxP-mediated functional ablation of the murine adar1 gene contained 6.9 kb of 5′-homology sequence including exons 3-6 and a pgk-neo gene flanked by two loxP511 elements (
) cloned into an SmaI site and 3.3 kb including exons 7-11 as 3′-homology sequence with a third loxP511 element inserted into an MscI site in intron 9 (Fig. 1). The linearized targeting vector was electroporated in R1 ES cells (
), and PCR-mediated screening of 540 G418 (aminoglycoside antibiotic)-resistant clones yielded one clone with a correctly targeted adar1 allele, as confirmed by Southern analyses. Chimeric mice generated by injection of the targeted ES cell clone into C57BL/6-derived blastocysts transmitted the loxP-flanked (floxed) adar1f7-9 allele through the germ line. The adar1Δ7-9 allele was generated by breeding adar1f7-9 mice with Cre-deleter mice (
). The modified loxP511 elements were used to permit allele-specific Cre-mediated recombination in the presence of an adar2 allele floxed with canonical loxP sites. The recombination efficiency of loxP511 in adar1+/f7-9 mice was comparable with that of the floxed adar2 allele.
Cre-positive adar1f7-9 mice were backcrossed to eliminate the cre transgene, and germ line excision of the floxed adar1 gene segment was verified by Southern analyses and sequencing of genomic DNA from adar1Δ7-9 mice.
The adar1Δ2-13 allele was created by replacing exons 2-13 of the adar1 gene with a pgk-neo gene, 1.1 kb of intron 1A generated by PCR as 5′-homology sequence, and a cloned 6-kb genomic fragment containing exons 14 and 15 as 3′-homology sequence. Electroporation of the linearized targeting vector into R1 ES cells and PCR-based screening of 1602 G418-resistant clones for homologous recombination events in the adar1 locus yielded three clones with a correctly targeted adar1Δ2-13 allele, as confirmed by Southern analyses. Two clones were injected into C57BL/6-derived blastocysts and contributed to viable chimeric offspring. Highly chimeric mice backcrossed to either NMRI or C57BL/6 wild-type mice transmitted the adar1Δ2-13 allele through the germ line.
Embryo Dissection and Genotyping—Embryos were derived from timed matings of adar1+/Δ7-9 or adar1+/Δ2-13 mice, with noon of the vaginal plug day defined as E0.5, and carefully staged by somite counts in combination with morphological criteria (
). Embryos were genotyped by multiplex PCR analysis of genomic DNA isolated from yolk sacs or tail snips with adar1 allele-specific oligonucleotide combinations: 5′-CTGTTGGGCCAGTGGTTCAGTAGTCC-3′ (adar1+ allele), 5′-GCGTACGTCGACGGTATCGA-3′ (adar1Δ7-9 allele), 5′-GCCACTTCTCCCTGACTCCTG-3′ (common downstream primer), 5′-CCTGCACATGCCGTGATCAGC-3′ (common upstream primer), 5′-GGGTGGACAAGTACCACCTG-3′ (adar1+ allele), and 5′-TAGATCTATAGATCTCTCGTGGGATC-3′ (adar1Δ2-13 allele).
5-HT2CReceptor Transcript Editing in Primary Neuronal Cultures—Single cell suspensions of four to eight brains from E12 embryos for each genotype were individually plated on poly-l-lysine (Sigma)-coated culture dishes and cultured for 3 weeks in Neurobasal medium (Invitrogen) containing B-27 supplement (Invitrogen) under appropriate conditions (37 °C, 5% (v/v) CO2, and >95% humidity). After culturing, cells were washed with phosphate-buffered saline, and total RNA was extracted with TriReagent (Molecular Research Center, Inc.) and reverse-transcribed into cDNA using random primers. PCR products containing the five editing sites were generated with 5-HT2C receptor cDNA-specific oligonucleotides (5′-GGCCAGCACTTTCAATAGTCGTG-3′ and 5′-CAATCTTCATGATGGCCTTAGTCC-3′) and sequenced directly with a nested primer (5′-GGCGAATTCCATTGCTGATATGCTGGTG-3′). The approximate extent of editing was derived from the different peak heights for the two purines occurring in the five editing positions in DNA sequence chromatograms.
Hematoxylin/Eosin Staining and Immunohistochemistry—Procedures were as described (
). Paraffin sections (5 μm) of embryos fixed in Bouin's solution (Sigma) were processed for histology and stained with Gill's hematoxylin 3 (Polysciences, Inc.) and eosin Y (Polysciences, Inc.). Sections processed for immunohistochemistry were incubated with antibody to TER-119 (3 μg/ml; BD Biosciences) or α-fetoprotein (4 μg/ml; Santa Cruz Biotechnology) and stained with biotin-conjugated IgG (2 μg/ml; Vector Labs, Inc.) directed against the primary antibody using a Vectastain Elite ABC kit (Vector Labs, Inc.). Nuclear counterstaining was carried out with Gill's hematoxylin 3.
Methylcellulose Culture—Single cell suspensions of yolk sac (104 cells), fetal liver (104 cells), and peripheral blood (105 cells) were cultured in duplicate in 0.9% (w/v) methylcellulose (M3234, Stemcell Technologies) supplemented with interleukin-3 (10 ng/ml) and erythropoietin (4 units/ml) to stimulate the growth of erythroid colonies or with 2% pokeweed mitogen murine spleen cell-conditioned medium to promote myeloid/granuloid colony growth. Colonies were counted 3 days (colony-forming unit-erythroid (CFU-E)), 5 days (burst-forming unit-erythroid (BFU-E)), and 9 days (colony-forming unit-myeloid/granuloid (CFU-M/G) post-seeding. For numbers of tissues analyzed, see Fig. 5B.
Flow Cytometry—Single cell suspensions (2 × 105 cells) of livers from two wild-type, two adar1+/-, and six adar1-/- E11.5 embryos were stained with fluorescein isothiocyanate-conjugated monoclonal antibody against c-Kit (clone Ack4, DPC Biermann), recorded in a FACSCalibur, and analyzed with CellQuest software.
Colony-forming Unit-Spleen (CFU-S) Analysis—Single cell suspensions of 61 aorta-gonad-mesonephros (AGM) regions and 62 fetal livers from wild-type E11 embryos and of 35 AGM regions and 37 fetal livers from adar1-/- E11 embryos (37-45 somite pairs) were injected into the lateral tail veins of sublethally irradiated (11.5-12.5 grays) adult female NMRI and C57BL/6 mice. Single macroscopic CFU-S colonies were isolated from the recipient spleens 11 days after injection, and genomic DNA was prepared. Male wild-type and mutant ADAR1 embryo-derived colonies were identified by PCR analysis with adar1 allele-specific (see above) and Y chromosome-specific (pYMT2/B; GenBank™/EBI accession number X05260; 5′-GTACTGGAGCTCTACAGTGATG-3′ and 5′-CAACACATCACTGTCATCTCCTG-3′) oligonucleotides.
adar1-/-ES Cells—An adar1+/Δ2-13 ES cell clone was plated and selected in high G418 concentrations (2.5 mg/ml) for cells with two adar1 null alleles. PCR-based screening of 240 G418-resistant colonies with adar1 allele-specific oligonucleotides (see above) yielded three adar1Δ2-13/Δ2-13 clones, as confirmed by Southern analyses.
Chimera Analysis—Chimeric mice were generated by injection of two independent adar1Δ2-13/Δ2-13 ES cell clones (129/Sv background) into wild-type blastocysts (C57BL/6 × NMRI background), followed by implantation in pseudopregnant foster mice. Genomic DNA was prepared from 15 tissues of four adult (postnatal day 50) chimeric mice, and the contribution of adar1-/- ES cells to these tissues was evaluated by two PCR-based analyses. An adar1 allele-specific PCR (primer sequences are given above) documented the presence of wild-type and null alleles. The extent of tissue colonization by the ADAR1-deficient ES cells was quantified from the sequences of PCR products containing a particular single nucleotide polymorphism (SNP; mm3_snpNih_rs3724839, the particular SNP in the SNP database of the National Institutes of Health) on chromosome 3 to distinguish the “parental” mouse strains (T for 129/Sv and C for NMRI and C57BL/6). PCR products containing the SNP were generated with oligonucleotides 5′-GGGGAAGGTGAAGTCACCACTG-3′ and 5′-CTCAAAGCTAGCACATGTTGTGCTG-3′ and sequenced directly with a nested primer (5′-CCCCTGGTTAGTGTGATAGAGTG-3′).
Different adar1 null alleles Produce the Same Embryonic Phenotype—We constructed a targeting vector for Cre/loxP-mediated functional ablation of the murine adar1 gene by introducing a pgk-neo gene flanked by two modified loxP elements (see “Experimental Procedures”) into intron 6 and a third such element into intron 9 of cloned adar1 genomic sequence (Fig. 1). The loxP-flanked (floxed) sequence encodes residues 707-870 of the interferon-inducible form of ADAR1 (GenBank™/EBI accession number AAK16102), including essential regions for editing activity in vitro (Fig. 1) (
) for germ line excision of the floxed gene segment. Heterozygous adar1+/Δ7-9 mice were phenotypically normal, but adar1+/Δ7-9 intercrosses produced no homozygous offspring among 119 pups (83 heterozygous and 36 wild-type), indicating interference of ADAR1 deficiency with embryonic development.
The normal appearance of adar1+/Δ7-9 mice was unexpected based on a report on lethality in chimeric embryos caused by a large population of cells heterozygous for a particular adar1 null allele (
) might indicate milder functional interference by our adar1Δ7-9 allele. To minimize the potential for interfering products, we constructed a new null allele (Fig. 1) in which exons 2-13, encoding residues 11-1054, which comprise all three dsRNA-binding domains, two putative Z-DNA-binding domains (
), and most of the catalytic domain of ADAR1, were replaced with a pgk-neo gene. Highly chimeric mice derived from two independent ES cell clones transmitted this severely truncated adar1Δ2-13 allele through the germ line and produced phenotypically normal heterozygous adar1+/Δ2-13 mice, in congruence with adar1+/Δ7-9 mice. In further congruence, no homozygous offspring came from adar1+/Δ2-13 intercrosses, whereas heterozygous and wild-type pups were born at expected frequency. Hence, differently configured adar1 null alleles cause embryonic death upon homozygosity, but not heterozygosity.
Selective Loss of RNA Editing by ADAR1 Deficiency—As ADAR1 is thought to be involved in site-selective A-to-I editing (
) was lost, but was spliced correctly after culturing neurons from E11.5 brains for 3 weeks. Use of this procedure was validated by our observation that cultured neurons from wild-type embryos exhibited extents of editing at the 25 known sites (not documented) that were generally within 10-20% of those determined for postnatal brain (
). ADAR1 deficiency had a remarkably selective effect at sites A and B, at which editing dropped from 80-90% in the wild type to undetectable levels, indicating that these sites are edited exclusively by ADAR1. Site C, only five nucleotides downstream of site B, was unaffected by loss of ADAR1. Editing at site C′, which is the least edited site in vivo (
), and site D was elevated relative to the wild type, which may reflect easier access by ADAR2 in the absence of ADAR1. Collectively, these findings identify site-selective A-to-I conversion in pre-mRNA as an in vivo function of ADAR1, previously only shown for ADAR2 (
). They further indicate that the two ADAR proteins edit distinct subsets of closely spaced adenosines.
We also studied the effect of dual ADAR deficiency on the editing sites in 5-HT2C receptor transcripts, again employing cultured neurons from E11.5 brains of the appropriate genotypes. Because ADAR2-deficient mice die within a few weeks after birth (
). Selective ADAR2 deficiency affected the extent of editing least at sites A and B, again confirming that these are edited by ADAR1. Importantly, we failed to detect any A-to-I conversion in 5-HT2C receptor transcripts in neurons with dual ADAR deficiency (Fig. 2); and hence, the combined activities of ADAR1 and ADAR2 mediate all known A-to-I editing in transcripts for this receptor.
ADAR1-deficient Embryos—To define the time and nature of the embryonic death resulting from loss of ADAR1, we inspected and genotyped embryos derived from both adar1+/Δ7-9 (data not shown) and adar1+/Δ2-13 (Table I) intercrosses at different developmental stages. We obtained indistinguishable results for both ADAR1-deficient mouse lines in two different strains (C57BL/6 and NMRI), hereafter referred to as adar1-/- mice. Viable adar1-/- embryos, defined as those with beating hearts at the time of dissection, were identified at mendelian frequency between E8.5 and E11.5. However, whereas most adar1-/- embryos (70%) were alive at E12.0, none survived E12.5. We conclude that ADAR1-deficient embryos die in utero between E11.5 and E12.5.
Beginning at E11.25-11.5 and as strikingly apparent by E12, adar1-/- embryos exhibited a significant reduction in fetal liver size, which preceded a progressively pale yolk sac and an often pale embryo proper compared with wild-type embryos (Fig. 3, A-C). The size of adar1-/- fetal livers was normal until E11.0-11.25, but did not increase further, whereas wild-type and adar1+/- livers gained ∼50% in size between E11.5 and E12.5. At this time, ADAR1-deficient fetal liver had an intact Glisson capsule, but became transparent. Shortly before death, adar1-/- embryos were slightly retarded in development (6-12 h), as shown by morphological criteria and somite numbers compared with age-matched wild-type embryos, although development appeared otherwise normal (Fig. 3B).
Microscopic examination of adar1-/- embryos (Fig. 3D) revealed in fetal liver from E11.5 onward a significantly reduced cell density and often blood accumulation in pathologically enlarged spaces, perhaps resulting from massive cell death. The examination failed to detect morphological abnormalities in other tissues, including the neuroepithelium, heart, branchial arches, placenta, and yolk sac. adar1-/- yolk sacs, even when pale, exhibited properly formed blood islands and vessels containing nucleated erythrocytes. Furthermore, staining of endothelial cells in adar1-/- yolk sacs with antibody to platelet-endothelial cell adhesion molecule-1 (
) indicated normal organization of the vasculature, and the vitellin vessels connecting the yolk sac and embryo proper appeared intact (data not shown). We did not find in adar1-/- embryos at E12 any sign of hemorrhaging into body cavities such as brain ventricles, pericardium, and peritoneum.
Cell proliferation in adar1-/- embryos was comparable with that in wild-type embryos at E11.0-11.25, including fetal liver, as revealed by bromodeoxyuridine incorporation. Slightly later, at E11.5, reduced numbers of bromodeoxyuridine-positive cells (30% of the wild type) were found in ADAR1-deficient fetal livers, although bromodeoxyuridine-positive cells were readily detectable in other tissues (data not shown).
Lack of Hematopoietic Cells in adar1-/-Fetal Livers—Around E11, the fetal liver becomes the major hematopoietic tissue, giving rise to the blood cell lineages found in the adult until, shortly before birth, hematopoiesis shifts to spleen and bone marrow (
). At E11.5 (but not yet at E10.5), cell numbers in ADAR1-deficient embryonic hematopoietic tissues were significantly reduced in the yolk sac (to 52 ± 8%, n = 19), fetal liver (to 42 ± 9%, n = 19), and peripheral blood (to 63 ± 9%, n = 11) compared with adar1+/- and wild-type embryos.
ADAR1-deficient fetal livers displayed from E11.25 onward a pronounced scarcity of hematopoietic cells with characteristically dark-staining nuclei, which were readily detectable in large numbers in wild-type liver (Fig. 3E). Moreover, many cells appeared with pyknotic nuclei (Fig. 3E) containing fragmented DNA, indicative of apoptotic cell death, as supported by TUNEL assays (data not shown). The majority of hematopoietic progenitors present in fetal liver around E11-13 differentiates into erythroid cells (
). Accordingly, we observed in wild-type liver, beginning at E11.25, a dramatic increase in the number of TER-119-expressing erythroid cells (Fig. 4). In contrast, only a few erythroid cells, most of them diagnosed as yolk sac-derived nucleated erythrocytes, were identified in ADAR1-deficient fetal livers at E11.25-11.5 (Fig. 4).
Early hepatogenesis appeared intact in adar1-/- embryos, as judged by normal numbers of hepatoblasts expressing α-fetoprotein at E11.0-11.25 (Fig. 4). Moreover, semiquantitative analysis by reverse transcription-PCR indicated that the expression of nine hepatic marker genes (
), including serum albumin, α-fetoprotein, the stress-signaling kinase SEK1/MKK4, and transcription factors Foxa1 (hepatocyte nuclear factor-3α) and hepatocyte nuclear factor-4α, was comparable with that in wild-type embryos at E12 (data not shown). However, the breakdown of the ADAR1-deficient fetal liver architecture beginning at E11.5 and our chimera analysis (see below) suggest that also hepatoblasts undergo cell death.
Analysis of Hematopoietic Progenitors—Fetal livers from E11.5 adar1-/- embryos contained, compared with those from adar1+/- and wild-type embryos, significantly fewer hematopoietic progenitors with characteristic expression of the receptor tyrosine kinase c-Kit (
), as measured by flow cytometric analyses (Fig. 5A). We therefore determined the numbers of clonogenic hematopoietic progenitors in the yolk sacs, fetal livers, and peripheral blood from E11.5 embryos, employing an in vitro colony-forming assay. In all three tissues tested, we observed in adar1-/- embryos a drastic reduction in colony numbers of BFU-E, CFU-E, and CFU-M/G progenitors compared with control embryos (Fig. 5B), indicating a defect in definitive hematopoiesis. By contrast, primitive erythropoiesis in the yolk sac, which provides the first blood cells of the developing embryo, did not critically depend on ADAR1, as CFU-E progenitor numbers (given as means ± S.D.) in the yolk sacs from E8.5 adar1-/- embryos (21 ± 13 colonies, n = 3) were comparable with those from wild-type embryos (13 ± 4 colonies, n = 7).
Although ADAR1 deficiency drastically reduced the numbers of definitive hematopoietic progenitors, it did not affect their capacity to differentiate, as judged from various stages of differentiating cells found in cytospin preparations from ADAR1-deficient fetal livers at E11. Compatible with this notion, ADAR1-deficient and wild-type livers at E10.5 and E11.5 exhibited comparable expression (analyzed by reverse transcription-PCR) of 17 hematopoietic marker genes (
), including embryonic βH1 and adult-type β-major globins, erythroid Kruppel-like factor, erythropoietin receptor, and myeloperoxidase (data not shown).
Analysis of an Early Hematopoietic Progenitor in Vivo—The severely reduced numbers of progenitors for erythroid as well as myeloid/granuloid cells in ADAR1-deficient hematopoietic tissues indicated that CFU-S progenitors, which can differentiate along both lineages (
). Thus, we injected single cell suspensions of AGM regions and fetal livers from E11 wild-type and mutant embryos into irradiated recipients, isolated spleen colonies 11 days later, and verified donor origin by PCR analysis. We detected 19 CFU-S in 61 AGM regions from heterozygous and wild-type (adar1+/+) embryos and 10 CFU-S in 62 fetal livers from adar1+/+ embryos. In contrast, none of 35 AGM regions and 37 fetal livers from adar1-/- embryos resulted in spleen colony growth. These results strongly suggest a requirement for ADAR1 in CFU-S progenitors in both AGM regions and fetal liver.
Chimera Analysis—To identify tissues that cell-autonomously require ADAR1 function, we generated chimeric mice by injection of ADAR1-deficient ES cell clones (129/Sv background) into wild-type blastocysts of mixed NMRI × C57BL/6 background. Four adult chimeras (postnatal day 50) were dissected, and the contribution of adar1-/- ES cells to different tissues was evaluated by two PCR-based analyses. adar1 allele-specific PCR documented the presence of wild-type and null alleles (Fig. 6A). Moreover, the extent of tissue colonization by the ADAR1-deficient ES cells was quantified from the sequences of PCR products containing an SNP (mm3_snpNih_rs3724839) to distinguish the parental mouse strains (T for 129/Sv and C for NMRI and C57BL/6) (Fig. 6B). As revealed by these analyses, adar1-/- cells readily contributed to 10 of 15 tissues analyzed, including brain, kidney, lung, and heart (contribution range, 3-36%; and average contribution in 10 tissues, 13%). However, we failed to detect a contribution of adar1-/- cells to all hematopoietic tissues (bone marrow, spleen, thymus, and blood), in agreement with our analysis of ADAR1-deficient hematopoietic progenitors. Remarkably, adar1-/- cells were also absent in the liver; and hence, the requirement for ADAR1 might transcend hematopoietic lineages.
ADAR1, Like ADAR2, Edits Transcripts in Vivo—Our genetic analysis of A-to-I editing in 5-HT2C receptor transcripts provided proof for the hypothesis that ADAR1 catalyzes A-to-I conversion in vivo. It further demonstrated that, within a narrow substrate domain, ADAR1 and ADAR2 edit different adenosines without cross-interference. Hence, ADAR1 and ADAR2 appear to exert non-redundant functions. This is also indicated by our observation that the phenotype of embryos deficient in both ADAR proteins was indistinguishable from that caused by ADAR1 deficiency alone, even though these enzymes are expressed early in development, with transcripts detectable already in ES cells (
J. C. Hartner, M. Higuchi, and P. H. Seeburg, unpublished data.
Such early expression seems to have no essential function, as severe phenotypic consequences of ADAR1 deficiency become apparent only after E10 (this work) and of ADAR2 deficiency not before juvenile stages (
No Haploinsufficiency for ADAR1—We investigated the physiological role of ADAR1 in mice and found that homozygosity (but not heterozygosity) for differently configured null alleles consistently caused embryonic lethality between E11.5 and E12.5, along with retarded development, impaired fetal liver structure, and definitive hematopoiesis. Thus, the claim of haploinsufficiency for ADAR1 (
) could not be reproduced and seems to have resulted from faulty expression of the manipulated adar1 allele, potentially leading to insufficient ADAR1 levels in the embryo. As ADAR proteins edit as dimers (
), reduction of ADAR1 levels should affect functional complex formation in a nonlinear fashion.
Requirement for ADAR1 in Hematopoietic Progenitors—Our data indicate a requirement for ADAR1 in AGM region and fetal liver hematopoiesis, as revealed by the absence of spleen colony-forming activity of ADAR1-deficient blood cell progenitors in both tissues. The failure of adar1-/- ES cells to contribute to any hematopoietic tissue in the adult mouse chimeras generated by us may well suggest that hematopoietic progenitors require ADAR1 in a cell-autonomous fashion. ADAR1 is essential particularly in the proliferation and/or survival of these progenitors, as revealed by in vitro colony-forming assays in combination with bromodeoxyuridine and TUNEL assays in adar1-/- fetal livers. However, the capacity of the remaining hematopoietic progenitors to differentiate appeared to be unaffected by loss of ADAR1, judging for example from the presence of differentiated cells in cytospin preparations of fetal liver. Moreover, primitive erythropoiesis in the yolk sac did not require ADAR1, with adar1-/- embryos exhibiting properly formed yolk sac blood islands, numerous nucleated erythrocytes, and approximately normal numbers of CFU-E progenitors. Thus, differential dependence on ADAR1 becomes an additional distinctive feature of primitive and definitive erythropoiesis.
No Known Mouse Mutation Phenocopies ADAR1 Deficiency—It is instructive to compare ADAR1 deficiency with other mutations affecting definitive hematopoiesis (
). Most bear no resemblance to the phenotype of adar1-/- embryos. Certain phenoparallels are, however, provided by two mutants deficient in transcription factors required for definitive hematopoiesis, but apparently dispensable for primitive erythropoiesis. These are c-Myb (
AML1, the DNA-binding subunit of core-binding factor, is essential for the generation of all definitive hematopoietic cells and appears to be involved in the budding of early progenitors from the hemogenic endothelium (
), whereas in ADAR1-deficient embryos at E11, erythroid and myeloid/granuloid progenitors are still detectable at low frequency. Hematopoietic progenitors are also observed at drastically reduced numbers in c-Myb-deficient embryos as a result of compromised proliferation and/or survival in fetal liver (
). This phenotype resembles that of adar1-/- embryos, although lethality develops somewhat later in ontogeny, and c-Myb deficiency, in contrast to ADAR1 deficiency, is not known to affect the fetal liver structure (
ADAR1 and Hepatic Cells—ADAR1-deficient fetal liver exhibited, beginning at E11.25-11.5, a massive cell loss, ultimately resulting in its disintegration. A pronounced loss of hepatoblasts (but no interference with hematopoiesis) occurs also in mid-gestation embryos deficient in the stress-signaling kinase SEK1/MKK4 (
). Hepatoblasts may therefore also require ADAR1, a notion reinforced by our limited chimera analysis, in which descendants of ADAR1-deficient ES cells were conspicuously lacking in adult liver, but were readily detected in 10 other tissues of ectoderm, endoderm, and mesoderm origins. Notably, the failure to contribute to adult liver is not a consequence of defective fetal liver hematopoiesis, as ES cells deficient in the transcription factor GATA-2, which is required for fetal liver and yolk sac hematopoiesis, extensively contribute to adult liver (
). Collectively, these findings and considerations would evince a cell-autonomous function of ADAR1 also in the hepatic lineage.
Embryonic Lethality Caused by ADAR1 Deficiency—Embryonic lethality at E11.5-12.5 caused by impaired definitive hematopoiesis is rendered unlikely because aml1-/- embryos lacking all definitive blood cells survive up to E13, and c-myb-/- embryos exhibiting severely reduced numbers of definitive hematopoietic progenitors die at E14-15 (
). Thus, impaired definitive hematopoiesis cannot alone explain the death of adar1-/- embryos. Possibly, the disintegration of the fetal liver architecture, as observed in sek1/mkk4-/- embryos, which die at E10-12 with pale yolk sacs (
), leads to interruption of the blood circulation, inevitably resulting in anoxia and death. Indeed, we observed in most (but not all) dying adar1-/- embryos an accumulation of blood in fetal liver spaces, which might cause the pale appearance of adar1-/- yolk sacs and embryos. Alternatively, defects not revealed in our study might underlie the embryonic death caused by ADAR1 deficiency, as recently demonstrated in retinoblastoma gene-deficient mice, which die of placenta failure (
Is the Severe Embryonic Phenotype Explained by Editing Deficiency?—Our study does not reveal the molecular cause for the catastrophic events that befall the developing embryo as a result of ADAR1 deficiency. Although the capacity of ADAR1 as an RNA-editing enzyme in vivo has now been demonstrated, no transcripts that need ADAR1-mediated A-to-I editing are known in the fetal liver. In fact, given the low editing status of most known edits in brain-expressed transcripts, the scenario whereby an unedited adenosine in a transcript sets in motion at remarkable speed the demise of the mid-gestation embryo appears unlikely. It is, however, not impossible, as exemplified by the single ADAR2-mediated edit critical for postnatal brain development (
). Other scenarios that posit additional functional roles of ADAR1, such as a more general involvement in RNA metabolism, are even more speculative. Clearly, interesting biology that needs elucidating is revealed by ADAR1 deficiency.
We thank Dr. Frank N. Single for help with gene targeting and ES cell culture; Dr. Jochen Klock for help with histology; Dr. Hanno Hock for interpretation of cytospin preparations; Christian Faul, Hasan Avci, Dominik Krilleke, Stephan Koenig, Sebastian Gornik, Kai Sona, and Horst Groβkurth for technical assistance; Christiane Brandel, Martina Herford, Axel Erhardt, and Margarita Pfeffer for animal care; and Drs. Rolf Sprengel, Pavel Osten, and Stuart H. Orkin for critically reading the manuscript.