A recessive screen for genes regulating hematopoietic stem cells

Derived from the Greek word for blood (αιμα, haima), hematopoiesis describes the generation of fully differentiated effector blood cells from multipotential, self-renewing hematopoietic stem cells (HSCs).¹  There are 2 waves of hematopoiesis in vertebrates (Figure 1A). The primitive wave is derived from hemangioblasts that form in the posterior primitive streak and migrate onto the yolk sac (YS) from embryonic day 7 (E7.0)^2,3  where they give rise to primitive erythroid progenitors (EryP) in association with endothelial and vascular smooth muscle cells. EryP proliferate and differentiate synchronously to generate large nucleated red cells, which enter the circulation as the heart begins to beat at approximately E9.0.^3,4  Platelets are also generated in the YS and enter the fetal circulation early, where they may or may not perform a classic hemostatic function.⁵  The definitive hematopoietic wave expands in the fetal liver (FL) from approximately E11.5, but HSCs are produced before this time in the ventral wall of the dorsal aorta and YS. Their potential to make at least 8 different mature blood cell lineages is not realized until the FL bud is formed from the gut tube and a suitable microenvironment is established for HSC seeding. From E11.5 to E15.5, there is a massive expansion of the FL, which is initially geared for production of definitive enucleated red blood cells to sustain rapid fetal growth (Figure 1A). From approximately E15.5, myeloid lineages are produced and hepatocytes begin to differentiate within the FL as it transforms from a hematopoietic organ into an adult liver.

We know much about the genes that are critical for definitive stem cell generation and differentiation from reverse genetic approaches in mice, primarily gene knockouts.⁶  Most of the essential transcription factors were initially discovered as oncogenes, as proteins that bind to important cis-regulatory elements in other genes, or by sequence homology.^1,7  Some factors play dual roles in primitive and definitive hematopoiesis, whereas others are relatively definitive-specific. c-Myb is an example of the latter,⁸  whereas Scl/Tal-1 and Gata2 play essential roles in both waves.^9,10 

Forward genetic screens using the mutagen ethylnitrosourea (ENU) offer several advantages over reverse genetics for gene discovery.¹¹  Most importantly, they are phenotype-driven and so make no presumptions about which genes are involved in a particular process. Resulting point mutations also mimic those that characterize the majority of human inherited and acquired genetic diseases, and point mutations can lead to hypomorphic or antimorphic alleles in addition to loss-of-function alleles. The ability of mis-sense mutations to radically affect proteome regulatory networks and subsequent phenotypes has been well demonstrated in both yeast¹²  and bacteria.¹³  Mammalian examples are also beginning to be described from ENU mouse screens,^14-16  including at the stem cell level.¹⁷  Gain-of-function (dominant mutations) and precise mutations in critical cis-regulatory elements are also possible outcomes from ENU screens. These are difficult to achieve using reverse engineering technologies without leaving small marks (such as LoxP sites) behind.

Several important genes for HSC and blood cell production have been discovered through ENU in zebrafish.¹⁸  Initial screens focused on embryonic anemia so were biased toward finding mutations in patterning or primitive hematopoiesis. Whereas some mutations were found in known hematopoietic genes,¹⁹  many more were in unexpected genes.^20,21  Our screen builds on numerous successful dominant or haploinsufficient hematopoietic^22,23  and immunologic^11,14  screens to identify ENU phenodeviants in mice. We describe 2 complementary fluorescence-activated cell sorter (FACS)–based assays to detect perturbations in subsets of definitive HSC/progenitors, red cells, and platelets in midgestation (E14.5) mouse embyros. We conducted our screens on fetal tissues^24,25  (liver and blood) rather than adult bone marrow to (1) minimize loss of lethal recessive phenotypes before birth, (2) access a hematopoietic organ (FL) containing quantifiable subsets of HSC/progenitors, (3) reduce time to screening of the animal, and (4) reduce project costs. To date, we have screened more than 1300 G3 embryos from 45 pedigrees for 9 hematopoietic parameters in parallel and identified 6 recessive mutations. We have given these pedigrees Australian Aboriginal names to reflect the Australian origin of the screen, and as acknowledgment that our Aboriginal peoples understand land sustainability just as stem cells maintain and repair adult tissues after damage or degeneration.

All 8 wild-type strains, C57BL/6J (B6), C57BL/10SnJ (B10), C57L/J, BALB/b, C3H/HeH, CBA/CaJ, FVB/NJ, and 129/SvEv, and ENU pedigrees (B6) were maintained at the Australian Phenomics Facility. The morning of vaginal plug observation was E0.5. ENU pedigree founder (G0) B6 male mice were treated weekly with 90 mg/kg ENU over 3 weeks. All mouse experiments were approved by the Institutional Review Board of Australian National University.

A full list of antibodies is available in supplemental data (available on the Blood Web site; see the Supplemental Materials link at the top of the online article). All cells were analyzed on an LSRII (BD Biosciences) and FACS data analyzed using FlowJo Version 9.1 software (TreeStar).

Whole blood (175 μL) from adult mice was collected into ethylenediaminetetraacetic acid tubes and run through an Advia 120/2120 Hematology System analyzer (Siemens AG). Cytospins were at 500g for 4 minutes in a Cytospin 3 (Shandon). Cells were fixed in 100% methanol and stained in May-Grünwald-Giemsa. Paraffin sections of spleen were stained with hematoxylin and eosin, Perl stain, or reticulin using standard protocols (supplemental data).

DNA from (B6 × B10) F2 embryos was analyzed using a panel of 80 B6/B10 single nucleotide polymorphisms (SNPs) that on average span the mouse genome at 20-Mb intervals. Full details of SNP marker design and genotyping are listed in supplemental data.

Sequencing of candidate genes was to locate the causal ENU B6 nucleotide substitution. DNA was prepared from an individual affected mouse and primers designed for candidate genes to amplify all exons plus or minus 15 bp to cover splice junctions. Primers are available on request. Amplicons were then sequenced by the Sanger method on a 3730xl capillary sequencer (Applied Biosystems). This automated platform used Big Dye Terminator chemistry Version 3.1 (Applied Biosystems). The raw trace files were analyzed using Lasergene software (DNASTAR) against B6 reference genome. Mutations were confirmed in a second affected person.

The E308G mutation on c-Myb that included an N-terminal HA tag was recreated using a polymerase chain reaction-based mutagenesis approach, after which the mutant was cloned into BamHI and XhoI sites of pcDNA3.1 (Invitrogen). Reporter and glutathione S-transferase (GST) pull-down assays to detect interaction between Myb and CBP KIX were performed as described in supplemental data.

Data were analyzed for significance between groups using a 2-tailed Student t test. Differences were considered significant at P < .05.

Our primary aim was to find recessive mutations causing abnormal definitive HSC production or turnover. A secondary aim was mutations that resulted in defective red cell and/or platelet production. Most of the knowledge about HSC frequencies and behavior has come from various assays using adult bone marrow. However, FL is a very rich source of HSCs that outcompete bone marrow stem cells in most scenarios.^26,27  To efficiently screen for long-term HSC (LT-HSC) and progenitor cell phenotypes in FL, we first set up a high-throughput FACS assay using wild-type C57BL/6J (B6) mice (Figure 1B). We chose this strain for mutagenesis because we have extensive experience in using it for ENU screens,^11,14  the mouse reference genome was built using B6 DNA making ENU-generated SNP detection easier, and many monoclonal antibodies have often been used specifically in B6. In particular, functional HSC assays, such as competitive repopulation assays, are easier using B6 congenic strains because of the availability of CD45 strain-specific antibodies.

Using an 8-color single-stain protocol, 6 functionally distinct hematopoietic stem and progenitor cell subsets, LT-HSCs, multipotent progenitors (MPPs), common lymphoid progenitors (CLPs), common myeloid progenitors (CMPs), granulocyte/macrophage progenitors (GMPs), and megakaryocyte/erythrocyte progenitors (MEPs), were quantified in parallel (Figure 1B). Self-renewing LT-HSCs were gated according to the 5-color cKit⁺Lin⁻Sca1⁺CD150⁺flk2⁻ cell surface phenotype, whereas MPPs were cKit⁺Lin⁻Sca1⁺CD150⁻flk2⁺.^26,27  The 4 subsets of lymphoid and myeloid progenitors were gated according to published protocols.^28,29 

By a separate 3-color FACS analysis and cell size measurement on a log scale, we also established a companion high-throughput screening assay for subsets of blood cells and platelets in fetal blood (FB) based on recently published work (Figure 1C).^3,5  YS-derived primitive red blood cells (RBCs) were identified based on their large size, nucleus (ie, Hoechst 33342 positivity), and expression of the cell surface marker TER119. FL-derived definitive RBCs were medium-sized and expressed TER119 but were enucleated (Hoechst 33342-negative). Finally, platelets were very small in size and stained for CD41 but not Hoechst 33342. Together, these 9 subsets of cells assayed in parallel provided a high-resolution snapshot of early hematopoiesis in both a primary tissue and the periphery (Figure 1A). The use of 2 complementary screening assays each with multiple phenotypic readouts also provided internal controls to minimize the identification of false positives.

We analyzed FL and FB samples from B6 embryos aged E12.5, E13.5, and E14.5 to examine the dynamics of stem and progenitor cell proliferation. A 5-fold increase in nucleated cells in the FL was observed with each developmental day (Figure 2A). Primitive RBCs formed the majority (∼ 80%) of the circulating FB at E12.5, a frequency that dropped by approximately 20% with each developmental day (Figures 1A, 2B). This is consistent with the gradual switch from primitive to definitive erythroid cells in the circulation.^3,4  We found that primitive erythroid cells underwent progressive enucleation in the circulation to greater levels than realized until recently,³  such that by E14.5 approximately 20% were enucleated (Figure 1C). The reduction in primitive RBCs was mirrored by a relative increase in definitive RBCs from approximately 5% at E12.5 to approximately 25% at E13.5 and approximately 50% at E14.5 (Figure 2B). By comparison, mean platelets counts of approximately 0.5% to 1% relative to RBCs were relatively constant across developmental days (Figure 2B), consistent with substantial platelet production from the YS as well as the FL.⁵ 

Frequencies (Figure 2C) and absolute numbers (Figure 2D) of the 6 HSC/progenitor cell subsets in FL were quantified across developmental days. MEP was the predominant progenitor population, with the E14.5 frequency progressing MEP > GMP > CMP > MPP > CLP > LT-HSC. LT-HSC, MPP, and CLP were very low in number at E12.5, composing only a few hundred cells but expanding to 5000 to 20 000 cells by E14.5.

Based on these data, the optimal developmental age to conduct our screen was E14.5, when the primitive and definitive RBC compartments in FB were at approximately 50% relative frequencies (Figures 1, 2), facilitating accurate quantification of both. In addition, mice harboring targeted deletions of genes essential for definitive hematopoiesis survive to approximately E14.5 because the primitive wave can sustain embryos until then.^8,30  Furthermore, we reasoned that loss-of-function or hypomorphic alleles of genes involved in both primitive and definitive hematopoieis, such as Epo, EpoR, or SCL/tal-1, might be viable or recently dead at E14.5.³¹  Lastly, E14.5 FL is large and easy to process, whereas older FL is harder to dissociate into single cells.

Because an outcross to a non-B6 inbred strain is required to map the causative ENU mutation underlying any phenotype of interest, we needed to ensure that the FACS assays were robust and identical in the mapping strain eventually chosen. Numerous phenotyping surveys encompassing a broad range of biologic disciplines have been conducted in inbred strains,^32,33  but none has specifically focused on the variability on HSC/progenitor cell surface profile³⁴  and frequency^35,36  across a comprehensive number of inbred mouse strains, particularly with recent advances in identified HSC markers.^26,27 

We analyzed 7 inbred strains (C57BL/10SnJ, C57L/J, BALB/b, C3H/HeH, CBA/CaJ, FVB/NJ, and 129/SvEv) alongside our ENU-mutagenized B6 stain for the baseline composition and enumeration of HSC/progenitor cell subsets in FL, and erythroid cells/platelets in FB. There were some major surprises. For example, a clear lack of Sca1 cell surface expression in BALB/c, C3H/HeH, and CBA/CaJ strains resulted in an inability to recognize and score LT-HSCs using Sca1 (Figure 3A; supplemental Figure 1A). FVB/NJ and 129/SvEv strains also showed a reduction in Sca1 expression relative to B6, C57BL/10SnJ, and C57L/J. One possible explanation for this apparent absence of Sca1 surface expression is that the epitope recognized by the Sca1 monoclonal antibody may have amino acid changes because of functionally conservative SNPs in non-B6 strains. Alternative antibodies raised against different epitopes or polyclonal antisera could give different results. Nevertheless, our choice of mapping strain to detect HSC phenodeviants was limited to B6-related and FVB/N strains. In addition, CLP frequency was similar to C57BL/6 levels in B6 variants and 129/SvEv mice, but all other strains had lower frequencies (Figure 3A). In addition, a shift in CD34 and FcγR cell surface staining resulted in ambiguous GMP frequencies for all but C57BL/10SnJ, whereas the CMP/MEP subsets were less variable across the strains tested (Figure 3A; supplemental Figure 1B).

FB was also tested (Figure 3B). Although precise developmental timing between mouse litters was one experimental variable that sometimes caused skewed ratios between circulating primitive and definitive RBCs (Figure 2), there was a distinct lack of TER119 expression on the surface of primitive RBCs of C3H/HeH and FVB/NJ strains (supplemental Figure 2). TER119 expression on definitive RBCs was more uniform across all mouse strains; a SNP in the gene encoding the surface protein recognized by the TER119 monoclonal antibody is thus unlikely (supplemental Figure 2).

Based on these observations, we outcrossed our ENU-mutagenized B6 pedigrees to the C57BL/10SnJ (B10) strain because it showed the least variability in hematopoietic profile relative to B6 across all parameters examined, an unsurprising finding given their common ancestry³⁷  and close position on the mouse phylogenetic tree.³⁸  Although this meant a reduction in polymorphisms between these 2 strains and mapping power,³⁹  we thought this was superseded by the need to reliably identify phenodeviants on outcrossing.

We chose a concentration of ENU based on many years of experience in the Australian Phenomics Facility. Previous experiments have determined that we generate approximately 30 mutations in coding genes in each founder G1 male using a standard 3 × 90 mg/kg ENU.⁴⁰  Using our FACS assays, we screened 1375 G3 embryos at E14.5 from 45 B6 ENU pedigrees and identified 6 strains with phenotypes in HSC/progenitors and/or early blood cell production. Supplemental Table 1 documents the first 1107 embryos and 34 pedigrees screened. Average litter size was 8 and the embryo resorption rate was approximately 15%. Some pedigrees showed a consistently higher rate of embryo resorption possibly indicative of either a dominant or recessive early embryonic lethal phenotype. We found 6 ENU pedigrees with reproducible defects defined as present in at least 3 independent litters. ENU11.21 (kandarra), ENU11.28 (booreana), and ENU11.211 (mulkirri) have been outcrossed and heritability confirmed. A further 3 pedigrees, ENU11.208 (chokerre), ENU11.235 (wonggan), and ENU11.312 (kamu), are in the process of outcrossing for mapping (Table 1).

The ENU strain booreana (from the Aboriginal word for white) was so named because embryo 8 from the first litter of pedigree ENU11.28 (Figure 4A) showed an expansion of LT-HSCs (Figure 4B,D). Booreana homozygous embryos (boo/boo) are indistinguishable from littermates by inspection (Figure 5A). However, boo/boo FL has an increased frequency of LT-HSC (4-fold), MPP (3-fold), CLP (3-fold), variable CMP numbers, but decreased GMP (3-fold) and MEP (2-fold; Figure 4B). That some cell subsets had expanded and others reduced strongly suggested a bona fide phenodeviant. The distinctive boo phenotype recurred in one embryo (no. 17) in the third litter of pedigree ENU11.28 (supplemental Figure 3). In this case, CMP numbers were normal, but the other subsets were like those in the founder embryo.

We outcrossed the heterozygous boo/+ G1 founder male from pedigree ENU11.28 to a wild-type B10 female and then established booreana as a heritable breeding line by intercrossing F1(B6xB10) siblings and testing with the same FL FACS screening protocol. Fulfilling our expectation from baseline phenotyping of non-B6 strains (Figure 3), the introduction of B10 genome did not suppress the booreana phenotype, which was observed at an expected Mendelian ratio (5/20; Figure 4C). There was a consistent increase in LT-HSC and CLP, with reduction in MEP and GMP. CMP were variable because of difficulties in gating these cells accurately in boo/boo embryos (supplemental Figure 5). FB from boo/boo embryos showed an expansion in their primitive RBC subset and dramatic reduction in their definitive RBCs (Figure 4E; supplemental Figure 4), consistent with the decreased MEP frequency. However, boo/boo embryos showed a striking 5- to 10-fold increase in circulating platelets (Figure 4E; supplemental Figure 4).

We generated a boo/boo DNA pool (n = 12) alongside an unaffected pool of ENU11.28 embryos (n = 20) and analyzed these against a panel of 80 B6/B10 SNPs to map the underlying booreana mutation to a 57-Mb region on chromosome 10 containing approximately 300 genes. The proto-oncogene c-Myb resides in the interval and was a strong candidate for mutation given previous reports describing hematopoietic defects in mice harboring c-Myb mutations.²²  Resequencing of all c-Myb exons identified an A-to-G transition at nucleotide 923 in exon 8 only in those embryos that displayed the characteristic booreana phenotype (Figure 5B). The DNA mutation results in a glutamic acid to glycine substitution at amino acid 308 (E308G) in the transactivation (TA) domain of c-Myb (Figure 5C). Interestingly, the c-Myb^E308G/E308G mutation is only 5 amino acids from another ENU mutant, c-Myb^M303V/M303V (also an A-to-G transition), which resulted multilineage hematopoietic abnormalities (see “c-Myb and hematopoiesis” below).²³ 

To test the biochemical function of c-Myb^E308G/E308G, we cloned the mutant cDNA into an expression vector and tested its ability to transactivate the myeloperoxidase gene promoter in reporter assays. Compared with wild-type c-Myb, the E308G mutation is completely inert (Figure 5G). There was slight residual function in the M303V TA domain mutation, suggesting that the E308G mutation is more severe. Because c-Myb has been reported to bind p300 and CBP via its TA domain,²³  we also asked whether the mutation in the acidic residue disrupted interaction. We found complete absence of binding of CBP to E308G and L302A mutations in GST pull-down assays in 293T cells (Figure 5I).

Like c-Myb^M303V/M303V mice, we found that 8-week old boo/boo mice had markedly increased platelets (P < .001), anemia (P = .003), and increased reticulocytes (P < .001; Figure 5D). Boo/boo homozygotes also have neutropenia (P = .01), which was not reported in c-Myb^M303V/M303V mice, suggesting some biochemical differences between these alleles. There were red cell abnormalities, including polychromasia and Howell-Jolly body inclusions (Figure 5E-F). There was moderate splenomegaly in boo/boo adults (254 ± 46 mg, n = 6) compared with heterozygous littermates (115 ± 15 mg, n = 5) because of expansion of the red pulp (Figure 6A-B). On the other hand, the white pulp/lymphocyte component of spleen was reduced in boo/boo mice. The overall splenic architecture was disorganized in many cases with pink amorphous material in the red pulp, increased iron staining, and markedly increased reticulin staining (Figure 6C-F).

Megakaryocytes in boo/boo mutants were also increased approximately 8-fold in the bone marrow and approximately 20-fold in the spleen as determined by CD41 staining (Figure 7). In addition to an increase in number, there was an increase in nuclear ploidy of boo/boo megakaryocytes as determined by DRAQ5 incorporation in CD41⁺ cells (Figure 7). Together, these changes are very reminiscent of human myeloproliferative neoplasms (MPNs) such as essential thrombocythemia (ET) with myelofibrosis (MF; see “Homozygous Booreana mice are a model for human essential thrombocythemia and myelofibrosis” below).

Splenic FACS analysis of booreana adult mice revealed markedly reduced B-cell lymphopoiesis in homozygous mutants (Figure 6G-H), as well as an increased frequency of CD71⁺/TER119⁺ cells (Figure 6I-J), Gr1^hi neutrophils, Gr1^int monocytes, and TER119⁺ erythroid cells (Figure 6K-L). Gr1^int cells were confirmed to be monocytes by coexpression of Ly6G (data not shown). The percentage of splenic CD3⁺ T cells was reduced to approximately 50% of wild-type littermates (Figure 6M-N), but absolute splenic T-cell numbers are probably close to equivalent given the 50% increase in spleen weight.

Forward genetics offers the potential to find novel genes involved in any biologic process if a robust assay is established for reliable detection of phenodeviants. Dominant and recessive screens using high-throughput blood cell counting assays have successfully identified mutations in many hematopoietic genes.^11,14,22-25  Some were well known from previous studies, whereas others were unexpected players. Embryonic recessive screens have simply relied on identification of visible phenodeviants.^24,25  To our knowledge, this is the first attempt to use quantitative FACS assays to detect perturbations in all the major definitive HSC/progenitor cell subsets. Such an undertaking would be difficult in adult mice, but our father-daughter mating strategy and fetal screen of G3s have enhanced the feasibility of such an undertaking; with a small team, we have performed a pilot screen of 45 pedigrees and found 6 recessive mutations. The rapid cloning of a novel point mutation in the c-Myb oncogene provides proof of principle that informative mutations in key definitive stem cell genes can be identified by using FACS and a recessive breeding strategy.

In the process of searching for an appropriate mapping strain, we uncovered differences in surface expression of stem cell markers. For example, Sca1 is not expressed on the cell surface of HSCs in several commonly used mouse strains, such as BALB/c, C3H, and CBA, and is weak in 129/Sv. This has implications for using Sca1 for the study of HSC biology in mice generated from gene-targeted ES cell lines from a 129/Sv genetic background. One possibly trivial explanation is that the epitope recognized by the Sca1 monoclonal antibody (D7) is slightly divergent in the non-B6 strains because of an SNP that results in a conservative amino acid change. Thus, Sca1 might be present in a variant form on HSCs in these strains. Further studies with different hybridomas or polyclonal antisera to Sca1 would be necessary to resolve this.

The TER119 red blood cell epitope does not seem to be present on embryonic red cells of some strains (C3H and FVB). TER119 is associated with glycophorin A, but the precise epitope has not been identified.⁴¹  Its absence in the primitive erythroid wave of certain strains is intriguing; it is present in the definitive wave of these strains so the genes coding for TER119 must be present in the genomes, and the SNP argument is not likely to be valid in this case. Our results suggest failure of TER119 detection in certain strains is either the result of a lack of a key regulatory protein in the primitive wave of these strains or of a partner protein or processing pathway component, which normally leads to stable protein accumulation at cell surface. These results suggest TER119 has limited utility for the study of primitive erythroid cell differentiation in certain mouse strains.

c-Myb is a well-known critical regulator of hematopoiesis and stem cell biology in other organ systems.⁴²  Gene knockout causes embryonic lethality by E15.5 from anemia.⁸  Several ENU c-Myb alleles have been discovered using screens to detect blood abnormalities. The M303V mutation impairs interaction with transcriptional coactivators, such as p300 and CBP.²³  Our mutation in an acidic residue of the TA domain leads to more severe transcriptional crippling and complete failure to interact with CBP. The phenotype of adult c-Myb^E308G/E308G mice is probably slightly more severe than the phenotype of c-Myb^M303V/M303V mice, consistent with the in vitro gene reporter data. Interestingly, a mutation in the KIX domain of p300, which disrupts interaction with c-Myb, leads to a similar phenotype, suggesting that the c-Myb-p300/CBP interaction is critical for aspects of hematopoiesis.⁴³  Thus, because the phenotype of the c-Myb knockout is much more severe than the 2 ENU strains, which disrupt p300/CBP cofactor binding, there must be additional critical functions for c-Myb in vivo that do not depend on traditional coactivator binding. For example, c-Myb might have additional key roles as a transcriptional repressor or work via additional protein interactions to direct gene expression or chromatin architecture.

Carpinelli et al performed a dominant screen for perturbations in platelet counts in adult mice.²²  They found 2 alleles of c-Myb, which they named plt3 and plt4 because they displayed mildly increased platelet numbers in adult heterozygotes. c-Myb^plt3/plt3 and c-Myb^plt4/plt4 homozygote mice have very high platelet numbers (∼ 4000 × 10⁶/mL) like c-Myb^E308G/E308G mice, and also mild anemia and reduced splenic B cells. The plt3 mutation resides in the DNA-binding domain of c-Myb, although it retains significant ability to transactivate a reporter gene so is likely to retain some DNA binding. Only a fraction of c-Myb^plt3/plt3 mice survive the perinatal period so the phenotype is more severe than the c-Myb^E308G/E308G or c-Myb^M303V/M303V phenotypes as one might expect for a DNA-binding mutation. On the other hand, c-Myb^plt4/plt4 mice survive at Mendelian ratios from a c-Myb^plt4/wt cross, so is less severe than the DNA-binding mutation in plt3. The plt4 mutation results in mutation D384V within the negative regulatory domain (NRD) domain of c-Myb. The in vivo interactions of this NRD domain are still incompletely understood and probably shed important knowledge about protein networks in which c-Myb operates.

c-Myb^E308G/E308G adult mice develop splenomegaly and a phenotype reminiscent of the human MPN, ET, and MF. This is perhaps not surprising because fibrosis in humans with these diseases is thought to result from excess production of cytokines, such as PDGF via increased megakaryocyte mass in bone marrow and spleen.⁴⁴  In approximately 50% of human cases, the underlying genetic defect in ET and MF is a point mutation in the JAK2 kinase (V617F), which leads to constitutive activity in the absence of stimulation from receptors, such as EpoR and MPL.⁴⁵  There are also reported activating mutations in the cytoplasmic domain of the MPL receptor (often at amino acid 515) in some cases of MF⁴⁶  and other described mutations, but in many cases of MPN the underlying mutation is unknown. Our results suggest that it might be worth resequencing c-Myb in cases of JAK2 V617F-negative and MPL515-wild-type ET and MF. In addition, c-Myb^E308G/E308G mice could be used to test new small molecules that inhibit JAKs or other signaling pathways that are overly active in MPN, such as PDGF receptor signaling in marrow fibroblasts.

We were able to identify the c-Myb mutation in booreana quickly because the boo/boo phenotype was distinctive and previous mutations and gene targeting of c-Myb had been reported.^8,23  However, identification of the causative mutation in our other pedigrees is taking longer for several reasons. First, there are limited SNPs between B6 and B10, so there is limited power to find meiotic recombinants in the intervals of interest. We have used newly described SNPs derived from deep sequencing of the B10 genome (Bruce Beutler, The Scripps Research Institute, written communication, September 2009) to limit the interval in mulkirri to 34 Mb on chromosome 7 and the interval in kandarra to 26 Mb on chromosome 2 (Table 1). However, these large intervals include 558 and 641 coding genes in the intervals of mulkirri and kandarra, respectively. Further meiotic recombinants will not help narrow the interval because the bottleneck is in SNPs, not recombination events. This is a consequence of using the B10 strain, but this is unavoidable because of the strain specificity of our assays. We have resequenced the coding exons of a few genes in these intervals, which have a history in the hematopoietic field and are yet to find the causative mutation in either strain. It is possible the causative mutations reside outside coding regions (eg, promoters or enhancers), but this has been described rarely and is not the experience of our colleagues.⁴⁰  We think it is more likely the genes of interest are novel with respect to hematopoiesis.

There has been a remarkably rapid advance in fast and cheap deep sequencing technologies, which will aid mutation identification. These technologies are being used in large-scale cancer resequencing projects but are also ideally suited to detecting ENU mutations. One advantage of using B6 mice for mutagenesis is the reference genome was generated using this strain, so unknown and confounding mouse strain-specific SNPs are less likely to be troublesome, and the ENU-induced mutation is more likely to be obvious. There are different approaches in using these technologies, which are much cheaper than deep sequencing of the whole mutant genome. We are currently using solution capture and resequencing of coding exons within the mulkirri and kandarra intervals as an efficient and relatively cost-effective way to find the causative mutations.⁴⁷ 

The authors thank Chris Goodnow and Ed Bertram for scientific advice and financial support, Harpreet Vohra and Mick Devoy for assistance with flow cytometry, and Shelley-Mae Bolton, Ryan Dunstan, and Nadiah Roslan for animal husbandry.

This work was supported by the National Health and Medical Research Council (C.J. Martin Fellowship; P.P.), the Leukemia Foundation (scholarship; D.R.P.), and the University of Queensland (Postdoctoral Fellowship; P.Y.).

This article is dedicated to the memory of Jared Franklin Purton (1976-2009).

Wellcome Trust

Contribution: P.P. conceived and designed the study, collected, assembled, analyzed, and interpreted the data, composed figures, wrote the manuscript, and gave final approval of manuscript; R.T., B.W., A.E.H., D.R.P., P.Y., S.O.C., and R.L. collected and assembled the data and gave final approval of the manuscript; T.J.G. analyzed and interpreted data, edited the manuscript, and gave final approval of the manuscript; and A.C.P. conceived and designed the study, analyzed and interpreted the data, composed figures, wrote the manuscript, and gave final approval of manuscript.

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

Correspondence: Peter Papathanasiou, Australian Phenomics Facility, Australian National University, Acton ACT 0200, Australia; e-mail: peter.papathanasiou@gmail.com; or Andrew C. Perkins, Institute for Molecular Bioscience, University of Queensland, St Lucia QLD 4067, Australia; e-mail: a.perkins@imb.uq.edu.au.