Hmgb3: an HMG-box family member expressed in primitive hematopoietic cells that inhibits myeloid and B-cell differentiation

Hematopoietic stem cells (HSCs) are a rare population of cells that predominantly reside within adult bone marrow. They are capable of the exponential proliferation and multilineage differentiation necessary for life-long replenishment of all blood cell types.¹  In order to maintain their numbers, HSCs undergo self-renewal, generating progeny that retain the same ability to proliferate and differentiate. The mechanisms by which HSCs make decisions to self-renew or differentiate are still unclear.

We have generated a library of cDNA sequences expressed in adult mouse bone marrow cells depleted of cells expressing lineage-specific surface proteins (Lin^–) and expressing high levels of the tyrosine kinase receptor c-kit (c-kit^HI), which we previously have reported are enriched for HSCs.^2,3  One of the sequences we identified corresponds to mouse Hmgb3, a recently discovered member of a superfamily of high-mobility group DNA-binding proteins.⁴ Hmgb3 was originally identified in an EST database,⁴  and the protein is 85%-89% homologous to Hmgb1 and 2. Human HMGB3 is 99% homologous to the mouse gene and has been mapped to X chromosome band q28. Of the 36 Hmgb3 ESTs, 31 were derived from embryonic tissue. Northern analysis demonstrated high levels of Hmgb3 mRNA in 18-dpc mouse embryos, but no detectable mRNA in adult brain, lung, liver, spleen, and kidney.⁴ 

The high mobility group superfamily is subdivided into 3 groups. HMGB1, HMGB2, and HMGB3 are included in the HMG-Box family.⁵  In vertebrates, HMGB1 and 2 are expressed in all cell types. Mammalian HMGB1 and 2 preferentially bind to bent DNA with no sequence specificity.^6,7  HMGB1 and 2 can bend linear DNA and appear to play a role in the formation of nucleoprotein complexes by altering chromatin structure to allow factor binding.^8,9  HMGB1 and 2 also have been shown to interact with DNA-binding proteins such as Hox and Oct, steroid hormone receptors, RAG1 and 2, Rel, and p53.¹⁰  HMGB1 and 2 can either activate or repress transcription. For example, HMGB1 can hinder the interaction between the TATA-binding protein and TFIIB, whereas HMGB2 stabilizes the TFIID-TFIIA complex.^11,12  Although nothing is known about the function of Hmgb3, the high homology between Hmgb3 and other HMG-Box family members suggests that Hmgb3 may share some of the same properties with Hmgb1 and 2.

In this report we show that Hmgb3 is expressed in HSCs, primitive progenitor cells, and differentiating erythroid cells in the adult mouse. Analysis of transgenic mice containing a GFP gene inserted into the Hmgb3 locus showed that Lin^–, c-kit^HI, Hmgb3⁺ cells were found to be enriched for HSCs, common lymphoid progenitors (CLPs), and common myeloid progenitors (CMPs).^13,14  Enforced expression of Hmgb3 in hematopoietic stem cells inhibited B-cell and myeloid differentiation. Our data demonstrate that Hmgb3 expression can act as a marker for long-term repopulating cells and that Hmgb3 functions to inhibit the myeloid differentiation of primitive hematopoietic cells.

Mouse bone marrow Lin^–, c-kit^HI cells were isolated as previously described.³  Briefly, mouse bone marrow harvested from femurs was incubated in a cocktail of rat anti–mouse monoclonal antibodies that recognized the cell surface markers CD4 (clone GK1.5), CD8 (53-6.7), B220 (6B2), Mac-1 (M1/70), Gr-1 (8C5), and Ter119 (Pharmingen, San Diego, CA). Bound cells were removed by magnetic beads coated with goat anti–rat IgG (BioMag beads, Qiagen, Valencia, CA). Lin^– cells were then incubated with biotinylated rat anti–mouse anti–c-kit antibody and subsequently stained with streptavidin-phycoerythrin (PE) (Caltag, Burlingame, CA). The use of a biotinylated c-kit antibody allows a greater separation of c-kit⁺ cells. All cell sorting for this and subsequent procedures was performed on a fluorescence-activated cell sorter (FACS) Vintage SE (Becton Dickinson, San Jose, CA) using 488-nm argon and 633-nm helium-neon lasers.

A 1.2-kb Hmgb3 cDNA was amplified from MEL cell RNA using the following primers: 5′-AAGCTGCACAG-GGCGAACAATACAGTCAGGA-3′ (GenBank AF022465, bp 7-28) and 5′-CCAAGCTT-ACTTCTCATGCCGGCCGACC-3′ (bp 1242-1261). The primers included PstI (5′) and HindIII (3′) sites to allow cloning into pBS KS+ (Stratagene, La Jolla, CA). The sequence of the amplified cDNA was verified by sequencing. The 1.2-kb Hmgb3 cDNA was excised as a PstI/KpnI fragment and cloned into pEGFP-N2 and pEGFP-C2 (Clontech, Palo Alto, CA) to generate Hmgb3-GFP genes, where the eGFP gene is fused in frame to the 3′ and 5′ ends of the Hmgb3 cDNA, respectively. Both fusion genes are expressed from the cytomegalovirus (CMV) promoter and contain the neomycin resistance gene. The Hmgb3-GFP plasmids and pEGFP-C2 were linearized with NdeI and electroporated into 32D cells using a BioRad GenePulser set at 0.6 kV/25 μFd. Transfected 32D cells were washed and transferred to fresh medium. Twenty-four hours later, G418 (500 mg/mL final active concentration) was added, and selection proceeded for 21 days. Fluorescent microscopy was performed using fluorescein isothiocyanate (FITC) and 4′6-diamidino-2-phenylindole 2HCl (DAPI) filters.

Northern blot analysis was performed as previously described.¹⁵  RNA was isolated using TriZol according to the manufacturer's instructions (Invitrogen, Carlsbad, CA). Sources of RNA included bone marrow, spleen, thymus, brain, heart, liver, lung, and kidney of adult C57BL/6 mice and cytotoxic T-lymphocyte (CTLL), mouse erythroleukemia (MEL), and 32D cell lines. The probe used for Northern analysis was a 526-bp BclI-HindIII fragment located within the 3′ untranslated region of the Hmgb3 cDNA.

Cells of erythroid (Ter119⁺), granulocytic (Gr-1⁺), monocyte/macrophage (Mac-1⁺), B-lymphoid (B220⁺), and T-lymphoid (CD4/8⁺) lineage were separated by staining with PE-conjugated monoclonal antibodies described in “Isolation of lin^–, c-kit^HI adult bone marrow cells” followed by FACS sorting. Platelets/megakaryocytes were separated by staining with rat anti–mouse monoclonal antibody for CD41 (Pharmingen, clone MWReg30, FITC conjugated). Mast cells were obtained by culturing whole mouse bone marrow for 3 weeks in murine interleukin-3 (IL-3) (PeproTech, Rocky Hill, NJ). Adherent cells were removed by serial passaging, and cell density was maintained at 2 × 10⁵ cells/mL. Lin^–, c-kit^–, Sca-1^–, and Lin^–, c-kit⁺, Sca-1⁺, IL-7Rα^– bone marrow cells were isolated by staining lineage-depleted bone marrow with allophycocyanin (APC)–conjugated anti–c-kit (2B8: Pharmingen), PE-conjugated anti–Sca-1 (E13-161.7: Pharmingen), and biotinylated anti–IL-7Rα (α734: eBioscience, San Diego, CA). Secondary staining was performed with streptavidin PE-Cy5 (eBioscience). Cells were sorted by FACS based on isotype controls. RNA was isolated from all cell types using TriZol.

Reverse-transcriptase polymerase chain reaction (RT-PCR) was performed as previously described.¹⁶  Primers used for PCR amplification of wild-type Hmgb3 cDNA were forward: 5′-CGCTGTGATTGACACATCTC-3′ (GenBank bp 668-687) and reverse 1: 5′-GTACAGTTGACAACTCAAGG-3′ (GenBank bp 1008-1027) (Figure 2B). Primers used for PCR amplification of Hmgb3-KI cDNA were the identical forward primer and reverse 2: 5′-TGATACAGTTTCAGGGCGCTC-3′ (GenBank bp 802-822) (Figure 4B). Amplification of β2-microglobulin cDNA (forward: 5′-TGCTATCCAGAAAACCCCTC-3′, reverse: 5′-GTCATGCTTAACTCTGCAGG-3′) served as a control. Amplifications of Hmgb3 and β2-microglobulin were performed in independent reactions for all lineage-positive samples and in duplex reactions for whole bone marrow, Lin^–, c-kit^–, Sca-1^–, and Lin^–, c-kit⁺, Sca-1⁺, IL-7Rα^– samples. For both conditions, limiting dilutions were used to ensure amplifications of Hmgb3 and β2-microglobulin were performed within the linear range. Reactions were performed for 32 cycles at 94°C for 15 seconds, 58°C for 15 seconds, and 72°C for 30 seconds. Hmgb3 mRNA levels were measured and normalized to β2-microglobulin by performing densitometry on relative band intensities captured on a Phosphor Screen and resolved on a Typhoon 8600 PhosphorImager (Amersham Pharmacia, Piscataway, NJ).

To perform in situ hybridization, Lin^–, c-kit^HI, c-kit^LO, and c-kit^NEG bone marrow cells were isolated as described in “Isolation of lin^–, c-kit^HI adult bone marrow cells”. The cells were suspended in collagen and applied to slides. In situ hybridization was performed by Molecular Histology (Rockville, MD) using ³⁵S-labeled sense or antisense RNA transcripts of the 526-bp fragment of the Hmgb3 3′ untranslated region used for Northern analysis as a probe. Silver grains were visualized using dark field microscopy and photography was dark and light field.

A 4.1-kb PstI–HpaI fragment containing a region starting 100 bp 5′ of exon 1 and ending 335 bp downstream of the stop codon (EMBL sequence: X chromosome, bp 54786765–54790795) was excised from a 5.5-kb fragment that encompasses the entire Hmgb3 locus. A 1.3-kb NotI–BamHI fragment containing an internal ribosomal entry site (IRES) linked to a green fluorescence protein (GFP) gene (IRES-GFP) was digested from pIRES2-EGFP (Clontech), subcloned into pLITMUS38 (New England Biolabs, Beverly, MA), and subsequently excised with EcoRV and SalI. A triple ligation containing the 4.1-kb PstI–HpaI Hmgb3 fragment, the 1.3-kb EcoRV-SalI IRES-GFP fragment, and PstI-SalI digested pBSII KS+ inserted the IRES-GFP gene into the 3′ untranslated region of Hmgb3. This fragment comprised the 5′ arm of the targeting construct and was excised as a NotI-XhoI fragment and cloned into the matching sites of plasmid AC 641. AC 641 contains a pgk-Neo marker, flanked by loxP sites, the remainder of exon 5, and additional 3′ sequence (EMBL: bp 54791080–54793856), and an HSV-TK gene.

The targeting vector was linearized with NotI and electroporated into TC1 embryonic stem (ES) cells (0.6 kV/25 μFd) as described previously.¹⁷  Correctly targeted cells were identified by Southern blot analysis using probes outside the targeting region. Three homologous recombination events were identified among 186 G418 and gancyclovir-resistant ES cell clones. Twelve to eighteen ES cells from clones 81 and 114 were injected into 3.5-day-old C57BL/6 blastocysts. A total of 6 germ line chimeras (4 from 81 and 2 from 114) were bred to 129/SvJ female mice, and F1 generation transgenic mice were identified by Southern blot analysis of tail DNA.

Bone marrow cells isolated from male Hmgb3-KI mice were sorted by FACS into GFP⁺ and GFP^– populations for semiquantitative duplex RT-PCR as outlined. Sorted cells also were cultured in complete methylcellulose media (Methocult GF M3434; Stem Cell Technologies, Vancouver, BC, Canada), and colony forming units–culture (CFU-C) were scored after 7 days in culture.

For isolation of common lymphoid progenitors,¹³  male Hmgb3-KI Lin^– bone marrow cells were stained with APC-conjugated anti–c-kit, PE-conjugated anti–Sca-1, and biotinylated anti–IL-7Rα. Secondary staining was performed with streptavidin PE-Cy5. Littermate control Lin^– bone marrow cells were stained with the recommended isotype antibodies. Lin^–, c-kit⁺, Sca-1⁺, IL-7Rα⁺ cells were sorted into GFP⁺ and GFP^– populations and cultured in MethoCult M3630 (Stem Cell Technologies) containing 10 ng/mL human IL-7 and supplemented with 100 ng/mL murine stem cell factor (SCF) (R&D Systems, Minneapolis, MN) and 20 ng/mL human Flt3L (PeproTech). CFU–pre-B colonies were scored after 7-10 days. As a control, the same cells were cultured in Methocult GF M3434 media, supplemented with 20 ng/mL Flt3L, and CFU-Cs were scored after 14 days.

For isolation of common myeloid progenitors,¹⁴  male Hmgb3-KI Lin^– bone marrow was stained with APC-conjugated anti–c-kit, biotinylated anti–Sca-1 biotinylated anti–IL-7Rα, and either PE-conjugated anti-FcγRII/III (2.4G2) or PE-conjugated anti-CD34 (RAM34) (Pharmingen). Secondary staining was performed with streptavidin PE-Cy5. Littermate control Lin^– bone marrow cells were stained with the recommended isotype antibodies. Lin^–, c-kit⁺, Sca-1^–, IL-7Rα^–, FcγRII/III^LO, or Lin^–, c-kit⁺, Sca-1^–, IL-7Rα^–, CD34⁺ were sorted into GFP⁺ and GFP^– populations. Each population was cultured in Methocult GF M3434 supplemented with 20 ng/mL human Flt3L. Granulocyte macrophage colony-forming unit (CFU-GM), erythroid burst-forming unit (BFU-E), and granulocyte-erythrocyte-megakaryocyte-macrophage colony-forming unit (CFU-GEMM) were scored after 7-14 days in culture. As a control, the same cells were cultured in MethoCult M3630 media, supplemented with murine SCF and human Flt3L, and CFU–pre-B colonies were scored after 7-10 days.

Lin^–, c-kit^HI cells were isolated from male Hmgb3-KI bone marrow as described and sorted by FACS into GFP⁺ and GFP^– populations. 30 000 Lin^–, c-kit^HI, GFP⁺ or 50 000 Lin^–, c-kit^HI, GFP^– cells were mixed with 1 × 10⁷ male C57BL/6 bone marrow cells injected into lethally irradiated (1000 Gy, Cs¹³⁷ source) male 129 × C57BL/6 F₁ mice (Jackson Laboratory, Bar Harbor, ME). Hemoglobin electrophoresis of peripheral blood proteins and Southern blot analysis of bone marrow, spleen, and thymus DNA were performed after 16 weeks as described.¹⁸ 

A 0.6-kb NcoI-BamHI Hmgb3 cDNA fragment, containing the entire Hmgb3 coding region (GenBank: bp 28-628), was cloned into the matching sites in the MFGs retroviral vector (provided by Dr H. Malech, Bethesda, MD) followed by the addition of a BamHI fragment containing an IRES-GFP cassette. A control murine stem cell virus (MSCV) vector containing an IRES-GFP cassette (MGirL22Y) was obtained from Dr Brian Sorrentino (St. Jude Children's Research Hospital).¹⁹  Both constructs were introduced into GP+E86 packaging cells by standard techniques, and high titer producer cell clones were isolated by limiting dilution.²⁰ 

Transduction of mouse bone marrow with either the Hmgb3-IRES-GFP or IRES-GFP retroviral vector was performed as described.²¹  Briefly, C57BL/6 mice were treated with 5-flurouracil (5-FU: 150 mg/kg, Sigma, St Louis, MO) 48 hours prior to bone marrow harvest. Bone marrow cells were plated at 5 × 10⁵ cells/mL in Dulbecco Modified Eagle Medium (Invitrogen Gibco) supplemented with 15% fetal calf serum (FCS), 4 mM L-glutamine, 20 ng/mL murine IL-3, 100 ng/mL murine IL-6 (PeproTech), and 100 ng/mL murine SCF. Bone marrow cells were collected 48 hours later and cocultured at the same concentration with either Hmgb3-GFP or IRES-GFP producer cells in the same media supplemented with 6 μg/mL polybrene (Sigma). The bone marrow cells were harvested 46 hours later and washed in Hanks Balanced Salt Solution (Invitrogen) supplemented with 2% FCS. To differentiate transduced cells in vitro, harvested cells were sorted based on GFP levels, and both GFP⁺ and GFP^– cells were plated in Methocult GF M3434 supplemented with 20 ng/mL human Flt3L at 1 × 10⁴ cells per plate. For long-term transplantation, 2 × 10⁶ transduced bone marrow cells were injected (unsorted) into the lateral tail vein of female genetically anemic WBB6F₁-W/W^V mice (Jackson Laboratories).

Thymocytes, splenocytes, bone marrow, and peripheral blood were collected 16 weeks after transplantation and stained with PE-conjugated rat anti–mouse antibodies to CD4, CD8, B220, Mac-1, Gr-1, and Ter119. Positive staining and presence of GFP protein was detected by FACS. Southern blot analysis was performed on DNA isolated from bone marrow, spleen, and thymus tissues using a probe that hybridized to a 750-bp BamHI-EcoRI region from the GFP gene.

The Hmgb3 locus encodes a 1.5-kb mRNA transcript (Figure 1A). The gene contains 5 exons; only exons 2-5 are translated.⁴  The protein sequence of Hmgb3 predicted that Hmgb3 should contain nuclear localization signals. We transfected 32D cells with constructs that express Hmgb3 fused with GFP at either the 5′ and 3′ end. (Figure 1B, top). Fluorescence microscopy demonstrated that GFP signals entirely overlap those resulting from DAPI staining, indicating that Hmgb3 is found only in the nucleus (Figure 1B, bottom). No differences were observed with the N-terminal and C-terminal constructs, and GFP was detected in both the nucleus and cytoplasm of cells transfected with the control CMV-GFP construct (data not shown).

Northern blot analysis demonstrated that higher levels of Hmgb3 mRNA were present in MEL and 32D cells than in CTLL cells (Figure 2A). In adult mice, there were higher levels of Hmgb3 mRNA in bone marrow than in other tissues, confirming previous findings.⁴  RT-PCR analysis of RNA from lineage-positive bone marrow cells demonstrated high levels of Hmgb3 mRNA in Ter119⁺ erythroid cells, and low or absent levels of Hmgb3 mRNA in all other lineage-positive hematopoietic cells (Figure 2B). Hmgb3 mRNA levels within different lineage-negative populations were measured by semiquantitative duplex RT-PCR (Figure 2C). We compared relative levels of Hmgb3 expression in whole bone marrow cells, Lin^–, c-kit^–, Sca-1^–cells, and Lin^–, c-kit⁺, Sca-1⁺, IL-7Rα^– cells, which are enriched for HSCs.¹³ Hmgb3 expression increased 3-fold in cells enriched for HSCs compared to whole marrow and 12-fold compared to Lin^–, c-kit^–, Sca-1^–cells. The localization of Hmgb3 expression to the most primitive hematopoietic cells within the Lin^– fraction was confirmed using in situ hybridization of an antisense riboprobe to Lin^–, c-kit^HI, Lin^–, c-kit^LO, and Lin^–, c-kit^NEG cells. Approximately 50% of Lin^–, c-kit^HI cells were positive for Hmgb3 expression, while 10% of Lin^–, c-kit^LO cells and 0% of Lin^–, c-kit^NEG cells were positive. No hybridization was detected with the sense probe (Figure 2D).

We modified the Hmgb3 locus to insert a GFP gene linked to an IRES in the 3′ untranslated region by homologous recombination in male 129/SvJ ES cells (Hmgb3-KI) (Figure 3A). Because the complete coding sequences for both Hmgb3 and GFP are now part of the same transcript, this strategy allows us to track Hmgb3 expression using GFP without affecting Hmgb3 protein function. Positive ES clones were identified by Southern blot analysis (Figure 3B). Hmgb3-KI mice are phenotypically normal. Since Hmgb3 is an X-linked gene, only male mice hemizygous for the insertion were used in subsequent experiments. FACS analysis performed on male thymocytes, splenocytes, and whole bone marrow demonstrated that 21% to 29% of CD4⁺ and CD8⁺ thymocytes expressed GFP but CD4⁺, CD8⁺, and B220⁺ splenocytes and CD4⁺, CD8⁺, B220⁺, Gr-1⁺, and Mac-1⁺ bone marrow cells were negative for GFP expression. GFP was expressed in 54% of Ter119⁺ bone marrow cells (Figure 3C). CD4⁺, CD8⁺, B220⁺, Gr-1⁺, Mac-1⁺, and Ter119⁺ peripheral blood cells did not express GFP (data not shown). FACS analysis of Lin^– bone marrow cells demonstrated that approximately 50% of Lin^–, c-kit^HI cells expressed GFP (Figure 4A).

Lin^–, c-kit⁺ cells from Hmgb3-KI mice were sorted based on GFP expression (Figure 4A). Duplex RT-PCR performed on GFP⁺ and GFP^– populations sorted from whole bone marrow demonstrated the presence of the Hmgb3 transcript only in those cells that are GFP⁺ (Figure 4B). However, the relative stability of Hmgb3 protein compared to GFP is unknown, so it is possible that GFP^– cells may contain Hmgb3 protein. Lin^–, c-kit⁺, GFP^– cells contained a 2.5-fold higher frequency of CFU-GM than Lin^–, c-kit⁺, GFP⁺ cells (P < .01, Figure 4C). To analyze HSC activity, Lin^–, c-kit^HI, GFP⁺, and Lin^–, c-kit^HI, GFP^– cells were mixed with male C57BL/6 competitor bone marrow before injection into lethally irradiated male 129 × C57BL/6F₁ recipients. Polymorphisms in the β-globin locus were used to differentiate between the contributions of Hmgb3-KI and C57BL/6 HSCs; the 129/SvJ strain carries the diffuse allele and the C57BL/6 strain carries the single allele.¹⁸  Sixteen weeks after transplantation, hemoglobin electrophoresis was performed on recipients' peripheral blood (Figure 4D). Despite the fact that larger number of Lin^–, c-kit^HI, GFP^– cells were competed against control cells, Lin^–, c-kit^HI, GFP^– cells contributed minimally to long-term repopulation. The smaller number of Lin^–, c-kit^HI, GFP⁺ competed against the same number of control cells contributed approximately 28% of the erythroid cells in repopulated mice. Southern blot analysis of bone marrow, spleen, and thymus DNA demonstrated similar results (data not shown).

Common lymphoid progenitors (CLPs) express the IL-7 receptor (IL-7Rα), and IL-7Rα expression can be used to discriminate between CLPs and IL-7Rα^– HSCs.^13,22  CLPs in Hmgb3-KI bone marrow were identified as Lin^–, c-kit⁺, Sca-1⁺, and IL-7Rα⁺ cells (0.01% of bone marrow cells) and sorted into GFP⁺ and GFP^– populations (Figure 5A).¹³  Lin^–, c-kit⁺, Sca-1⁺, and IL-7Rα⁺ cells did not form myeloid colonies in culture but were highly enriched for CFU–pre-B activity (Table 1). Approximately 27% of all Lin^–, c-kit⁺, Sca-1⁺, and IL-7Rα⁺ cells were GFP⁺ and GFP⁺ cells displayed a 5-fold increase (P < .001) in the frequency of CFU–pre-B and a nearly 2-fold increase in absolute CFU–pre-B numbers compared to GFP^– cells (Figure 5B; Table 1).

Primitive myeloid progenitors express neither Sca-1 nor IL-7Rα.¹⁴  These progenitors in Hmgb3-KI bone marrow were identified as Lin^–, c-kit⁺, Sca-1^–, and IL-7Rα^– cells. Lin^–, c-kit⁺, Sca-1^–, IL-7Rα^– cells only formed myeloid colonies in culture; no CFU–pre-B were observed (data not shown). Lin^–, c-kit⁺, Sca-1^–, IL-7Rα^– progenitors can be further separated into distinct populations by staining with antibodies to FcγRII/III and CD34. The common myeloid progenitor (CMP) exhibits a FcγRII/III^LO and CD34⁺ phenotype. CMPs can differentiate into one of 2 lineage-restricted progenitors: the megakaryocyte-erythrocyte progenitor (MEP: FcγRII/III^LO and CD34^–) and the granulocyte-monocyte progenitor (GMP: FcγRII/III^HI and CD34⁺).

To determine if Hmgb3 expression could be used to separate CMPs from MEPs and GMPs, we sorted Lin^–, c-kit⁺, Sca-1^–, IL-7Rα^– cells from Hmgb3-KI bone marrow based on GFP and either FcγRII/III or CD34 expression. Lin^–, c-kit⁺, Sca-1^–, IL-7Rα^– FcγRII/III^LO cells (0.008% of bone marrow cells), which contain both CMPs and MEPs but not GMPs, were sorted into GFP⁺ and GFP^– populations (Figure 6A).¹⁴  Approximately 89% of Lin^–, c-kit⁺, Sca-1^–, IL-7Rα^–,FcγRII/III^LO cells expressed Hmgb3. Both FcγRII/III^LO GFP⁺ and FcγRII/III^LO GFP^– cells contained myeloid colony-forming cells of all lineages. The frequency of CFU-GM was 2-fold higher, and the frequency of CFU-GEMM was 3-fold higher in FcγRII/III^LO GFP^– cells (P < .01) compared to GFP⁺ cells (Figure 6B). Conversely, the frequency of BFU-E was 3-fold higher (P = .001) in FcγRII/III^LO GFP⁺ cells compared to FcγRII/III^LO GFP^– cells. The absolute number of colony-forming cells (CFCs) was 7-fold higher in FcγRII/III^LO GFP⁺ cells compared to FcγRII/III^LO GFP^– cells (Table 1).

Lin^–, c-kit⁺, Sca-1^–, IL-7Rα^– CD34⁺ cells (0.008% of bone marrow cells), which contain CMPs and GMPs but not MEPs, were sorted into GFP⁺ and GFP^– populations (Figure 6A)¹⁴ ; 75% of Lin^–, c-kit⁺, Sca-1^–, IL-7Rα^– CD34⁺ cells expressed Hmgb3 (Figure 6A). Both CD34⁺ GFP⁺ and CD34⁺ GFP^– cells contained myeloid colony-forming cells of all lineages. The frequency of CFU-GM was identical in both populations, and the frequency of CFU-GEMM was 2-fold higher in CD34⁺ GFP^– cells compared to CD34⁺ GFP⁺ cells (P < .05). Conversely, the frequency of BFU-E was 2-fold higher (P < .01) in CD34⁺ GFP⁺ cells compared to CD34⁺ GFP^– cells. The absolute number of CFCs was 3-fold higher in CD34⁺ GFP⁺ cells compared to CD34⁺ GFP^– cells (Table 1).

We investigated the effects of enforced expression of Hmgb3 by transducing bone marrow obtained from 5-FU–treated C57BL/6 mice with either a control vector containing an IRES-GFP cassette or an experimental vector containing an Hmgb3-IRES-GFP cassette (Figure 7A). Transduction efficiency was equivalent between the 2 vectors. Transduced marrow was sorted into GFP⁺ and GFP^– populations and plated in methylcellulose. Colonies of all lineages formed from both GFP⁺ and GFP^– cells transduced with the control vector and GFP^– cells transduced with Hmgb3-IRES-GFP. However, GFP⁺ cells transduced with Hmgb3-IRES-GFP were unable to form colonies (Figure 7B). To determine the effects of Hmgb3 overexpression in vivo, transduced (but unsorted) bone marrow was transplanted into genetically anemic WBB6F₁-W/W^V mice.²³  Bone marrow cellularities and peripheral blood counts of animals who received transplants of bone marrow cells transduced with either vector were equivalent. Sixteen weeks after transplantation, FACS analysis was performed on thymocytes, splenocytes, bone marrow, and peripheral blood. Representative results are shown in Figure 7C. In mice who received transplants of bone marrow transduced with the control IRES-GFP vector, GFP was detected in 10% to 62% of CD4⁺ and CD8⁺ thymocytes, B220⁺ splenocytes, Mac-1⁺, Gr-1⁺, Ter119⁺, and Lin^– bone marrow cells (Figure 7C). GFP also was expressed in 10% to 71% of all lineages of peripheral blood cells. In mice who received transplants of Hmgb3-IRES-GFP–transduced bone marrow, GFP was detected in approximately 30% of CD4⁺ and CD8⁺ thymocytes and 27% of Lin^– bone marrow cells, but GFP was detected in fewer than 1% of B220⁺ splenocytes, Mac-1⁺, Gr-1⁺, and Ter119⁺ bone marrow cells, and peripheral blood cells (Figure 7C). Southern blot analysis demonstrated many common integration sites for the control IRES-GFP vector in genomic DNA isolated from bone marrow, thymus, and spleen of repopulated mice (Figure 7D). In contrast, integrated GFP sequences were detected only in the thymus and occasionally the spleen DNA in mice transplanted with Hmgb3-IRES-GFP transduced marrow, but no GFP sequences were detected in bone marrow DNA.

Hmgb3 is a member of a family of chromatin-binding proteins that was originally discovered as an EST preferentially expressed in embryonic tissues.⁴  We have shown that Hmgb3 mRNA is present at relatively high levels in adult mouse bone marrow erythroid and primitive progenitor cells. HMGB1 and 2 have been shown to be expressed in all types of cells, but Hmgb3 represents the first example of an HMG-Box family member expressed in specific populations of hematopoietic cells.²⁴ 

Despite the fact that the relative stability of Hmgb3 protein compared to GFP is unknown, the data from our Hmgb3-KI mouse indicate that Hmgb3 expression can serve as a marker for most HSCs and CLPs. The pattern of Hmgb3 expression in myeloid progenitors is more complex. Both Hmgb3⁺ and Hmgb3^– myeloid progenitors gave rise to CFU-GM, BFU-E, and CFU-GEMM indicating that Hmgb3 expression is not confined to one myeloid lineage or one progenitor type and therefore is not a specific marker of CMPs, GMPs, or MEPs. However, in terms of absolute numbers, more myeloid progenitors express Hmgb3 than do not. The greatest differences in both CFC frequencies and absolute numbers were observed for BFU-E. Hmgb3 is expressed in a subset of Ter119⁺ bone marrow cells, and Hmgb3⁺ progenitors were enriched for BFU-E. This was observed in both FcγRII/III^LO progenitors (which contain CMPs and MEPs) and CD34⁺ progenitors (which contain CMPs and GMPs). We speculate that the Lin^–, c-kit⁺, Sca-1^–, IL-7Rα^–, FcγRII/III^LO, Hmgb3⁺ population significantly overlaps with MEPs (Lin^–, c-kit⁺, Sca-1^–, IL-7Rα^–,FcγRII/III^LO, CD34^–). Similarly, the Lin^–, c-kit⁺, Sca-1^–, IL-7Rα^–, CD34⁺, Hmgb3⁺ population, which does not contain MEPs, significantly overlaps with CMPs (Lin^–, c-kit⁺, Sca-1^–, IL-7Rα^–, FcγRII/III^LO CD34⁺) that have begun erythroid differentiation.

We propose that Hmgb3 levels must drop to allow normal differentiation program of myeloid progenitors and B cells because enforced overexpression of Hmgb3 inhibited myeloid and B-cell differentiation. The exact stage at which myeloid differentiation is affected by the level of Hmgb3 is not clear. It is possible that it varies between different myeloid lineages, since among lineage-positive myeloid cells in the bone marrow, only Ter119⁺ cells expressed Hmgb3. However, it also is possible that endogenous Hmgb3 expression is suppressed at the same stage in all myeloid cells, at which point Hmgb3 overexpression inhibits differentiation, and that Hmgb3 is subsequently re-expressed in Ter119⁺ bone marrow cells. Another possibility is that Hmgb3 is subject to posttranscriptional regulation, similar to SCL in erythroid cells, so that mRNA levels stay constant or even increase as protein levels decrease.²⁵ 

The observations that CD4⁺ and CD8⁺ thymocytes from Hmgb3-KI mice express Hmgb3 but B220⁺ cells in the spleen and bone marrow do not suggest that Hmgb3 may play a role in the specification of the B-cell lineage in CLP. Both HMGB1 and HMGB2 have been shown to interact with the RAG1 homeodomain and enhance recombination in vitro.²⁶  Given the high degree of homology between HMG-Box family members, it is possible that Hmgb3 can perform a similar function, critical in B-lymphocyte development.

HMG-Box family proteins bind to chromatin in a sequence-independent manner and have been shown to interact with several different types of DNA-binding proteins.¹⁰  Of these, the Hox and Oct family members both have been implicated in hematopoiesis.²⁷  HMGB1 has been shown to interact with HOXB1, which is expressed in more differentiated hematopoietic cells, and HOXB3, which has been found in CD34⁺, CD38^–cells.^28,29  Although it is unknown whether Hmgb3 can interact with Hox proteins, given the homology between Hmgb1 and 3, we speculate that Hmgb3 can interact with Hox proteins in primitive blood cells. Enforced expression of HOXB6, HOXB7, and HOXB8 in hematopoietic cell lines has been shown to block myeloid differentiation.^30-33  We propose that Hmgb3 overexpression may enhance recruitment of these Hox proteins to their cognate binding sites. HoxB6 expression has been demonstrated to play a role in erythropoiesis and the regulation of early erythrocyte progenitors.^34,35  The coexpression of Hmgb3 and HoxB6 suggests a cooperative role for these proteins in erythropoiesis.